index
stringlengths 21
21
| question
stringlengths 87
406
| A
stringlengths 3
170
| B
stringlengths 2
280
| C
stringclasses 1
value | D
stringclasses 1
value | bounding_box
listlengths 1
3
| direction
listlengths 0
3
| answer
stringclasses 4
values | answer_cot
stringlengths 385
3.54k
| answer_name
stringlengths 6
164
| category
stringclasses 12
values | image_url
stringlengths 20
20
|
|---|---|---|---|---|---|---|---|---|---|---|---|---|
0047fa2c7d9bfd25_dcd7
|
Consider the real-world 3D locations and orientations of the objects. Which object is a wooden chair with a brown seat facing towards, a wooden chair with a white cushion or the a black poster with a woman on it?
|
a wooden chair with a white cushion
|
a black poster with a woman on it
| null | null |
[
{
"bbox_3d": [
1.3,
0.5,
7.4
],
"label": "a wooden chair with a brown seat"
},
{
"bbox_3d": [
0.7,
0.4,
7
],
"label": "a wooden chair with a white cushion"
},
{
"bbox_3d": [
0.7,
0.9,
2.4
],
"label": "a black poster with a woman on it"
}
] |
[
{
"front_dir": [
-0.1,
0,
-1
],
"label": "a wooden chair with a brown seat",
"left_dir": [
-1,
0,
0.1
]
},
{
"front_dir": [
0.4,
-0.1,
-0.9
],
"label": "a wooden chair with a white cushion",
"left_dir": [
-0.9,
0,
-0.4
]
}
] |
B
|
To solve this problem, we first detect the 3D location of a wooden chair with a brown seat, a wooden chair with a white cushion, and a black poster with a woman on it. Then we compute the cosine similarities between the front direction of a wooden chair with a brown seat and the vectors from a wooden chair with a brown seat to the other two objects. We can estimate the angles from the cosine similarities, and the object with a smaller angle is the object that a wooden chair with a brown seat is facing towards. The 3D location of a wooden chair with a brown seat is (1.3, 0.5, 7.4). The 3D location of a wooden chair with a white cushion is (0.7, 0.4, 7.0). The 3D location of a black poster with a woman on it is (0.7, 0.9, 2.4). The front direction of a wooden chair with a brown seat is (-0.1, 0.0, -1.0). First we consider if a wooden chair with a brown seat is facing towards the a wooden chair with a white cushion. The vector from a wooden chair with a brown seat to a wooden chair with a white cushion is (-0.6, -0.1, -0.4). The cosine similarity between the front direction and the vector is 0.59, corresponding to an angle of 54.02 degrees. First we consider if a wooden chair with a brown seat is facing towards the a black poster with a woman on it. The vector from a wooden chair with a brown seat to a black poster with a woman on it is (-0.6, 0.4, -4.9). The cosine similarity between the front direction and the vector is 1.00, corresponding to an angle of 4.74 degrees. We find that the angle between the front direction and a black poster with a woman on it is smaller. Therefore, the final answer is B. a black poster with a woman on it.
|
B. a black poster with a woman on it.
|
multi_object_facing
|
0047fa2c7d9bfd25.jpg
|
0048f11fc5b63d5f_a4e9
|
Consider the real-world 3D locations and orientations of the objects. If I stand at a blue sign with a white circle in the middle's position facing where it is facing, is a man in red hat walking in front of me or behind me?
|
in front of
|
behind
| null | null |
[
{
"bbox_3d": [
0.7,
0.8,
2
],
"label": "a man in red hat walking"
},
{
"bbox_3d": [
1.3,
2,
4.4
],
"label": "a blue sign with a white circle in the middle"
}
] |
[
{
"front_dir": [
0.3,
0,
-1
],
"label": "a blue sign with a white circle in the middle",
"left_dir": [
-0.9,
-0.4,
-0.3
]
}
] |
A
|
To solve this problem, we first determine the 3D locations of a man in red hat walking and a blue sign with a white circle in the middle. Then we estimate the vector pointing from a blue sign with a white circle in the middle to a man in red hat walking, as well as the front direction of a blue sign with a white circle in the middle. Next we compute the cosine similarities between the vector and the front direction, which leads to the angle between the vector and the fron direction. If the angle is smaller than 90 degrees, then a man in red hat walking is in front of a blue sign with a white circle in the middle. Otherwise, a man in red hat walking is behind a blue sign with a white circle in the middle. The 3D location of a man in red hat walking is (0.7, 0.8, 2.0). The 3D location of a blue sign with a white circle in the middle is (1.3, 2.0, 4.4). The vector from a blue sign with a white circle in the middle to a man in red hat walking is hence (-0.6, -1.2, -2.4). The front direction of a blue sign with a white circle in the middle is (0.3, 0.0, -1.0). The cosine similarity between the vector and the front direction is 0.78, corresponding to an angle of 38.90 degrees. The angle is smaller than 90 degrees, meaning that a man in red hat walking is in front of a blue sign with a white circle in the middle. Therefore, the final answer is A. in front of.
|
A. in front of.
|
orientation_in_front_of
|
0048f11fc5b63d5f.jpg
|
004907203d47e92d_cb3f
|
Consider the real-world 3D locations and orientations of the objects. Which side of a white laptop with a black microphone is facing the camera?
|
front
|
left
|
back
|
right
|
[
{
"bbox_3d": [
-0.2,
0.2,
0.9
],
"label": "a white laptop with a black microphone"
}
] |
[
{
"front_dir": [
0.3,
0,
-0.9
],
"label": "a white laptop with a black microphone",
"left_dir": [
-0.9,
0.1,
-0.3
]
}
] |
A
|
To solve this problem, we first estimate the 3D location of a white laptop with a black microphone. Then we obtain the vector pointing from the object to the camera. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a white laptop with a black microphone, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a white laptop with a black microphone that is facing the camera corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The location of a white laptop with a black microphone is (-0.2, 0.2, 0.9). The vector from a white laptop with a black microphone to camera is hence (0.2, -0.2, -0.9). The left direction of a white laptop with a black microphone is (-0.9, 0.1, -0.3). The cosine similarity between the vector pointing to camera and the left direction is 0.06, corresponding to an angle of 86.43 degrees. Thus the angle between the vector pointing to camera and the right direction is 93.57 degrees. The front direction of a white laptop with a black microphone is (0.3, -0.0, -0.9). The cosine similarity between the vector pointing to camera and the front direction is 0.98, corresponding to an angle of 10.84 degrees. Thus the angle between the vector pointing to camera and the back direction is 169.16 degrees. Among the four directions, the smallest angle is the front direction, with an angle of 10.84 degrees. Thus the front side of a white laptop with a black microphone is facing the camera. Therefore, the final answer is A. front.
|
A. front.
|
orientation_viewpoint
|
004907203d47e92d.jpg
|
00493f6c61b373d3_eecb
|
Consider the real-world 3D locations and orientations of the objects. If I stand at a pink car with a license plate's position facing where it is facing, is a red car on the left or right of me?
|
on the left
|
on the right
| null | null |
[
{
"bbox_3d": [
-2.3,
0.7,
5.4
],
"label": "a red car"
},
{
"bbox_3d": [
0.6,
0.8,
3.3
],
"label": "a pink car with a license plate"
}
] |
[
{
"front_dir": [
0.2,
-0.1,
-1
],
"label": "a red car",
"left_dir": [
-1,
0,
-0.2
]
},
{
"front_dir": [
-0.5,
0,
-0.8
],
"label": "a pink car with a license plate",
"left_dir": [
-0.8,
0,
0.5
]
}
] |
A
|
To solve this problem, we first determine the 3D locations of a red car and a pink car with a license plate. Then we estimate the vector pointing from a pink car with a license plate to a red car, as well as the left direction of a pink car with a license plate. Next we compute the cosine similarities between the vector and the left direction, which leads to the angle between the vector and the fron direction. If the angle is smaller than 90 degrees, then a red car is on the left of a pink car with a license plate. Otherwise, a red car is behind a pink car with a license plate. The 3D location of a red car is (-2.3, 0.7, 5.4). The 3D location of a pink car with a license plate is (0.6, 0.8, 3.3). The vector from a pink car with a license plate to a red car is hence (-2.8, -0.0, 2.0). The left direction of a pink car with a license plate is (-0.8, 0.0, 0.5). The cosine similarity between the vector and the left direction is 1.00, corresponding to an angle of 3.68 degrees. The angle is smaller than 90 degrees, meaning that a red car is on the left of a pink car with a license plate. Therefore, the final answer is A. on the left.
|
A. on the left.
|
orientation_on_the_left
|
00493f6c61b373d3.jpg
|
0049a8ab0b787a1a_54a0
|
Consider the real-world 3D locations and orientations of the objects. If I stand at a bulletin board with a man on it's position facing where it is facing, is a wine glass with a yellowish tint in front of me or behind me?
|
in front of
|
behind
| null | null |
[
{
"bbox_3d": [
0,
0.2,
1.1
],
"label": "a wine glass with a yellowish tint"
},
{
"bbox_3d": [
0.6,
0.8,
1.5
],
"label": "a bulletin board with a man on it"
}
] |
[
{
"front_dir": [
-0.5,
-0.1,
-0.9
],
"label": "a bulletin board with a man on it",
"left_dir": [
-0.9,
0.1,
0.4
]
}
] |
A
|
To solve this problem, we first determine the 3D locations of a wine glass with a yellowish tint and a bulletin board with a man on it. Then we estimate the vector pointing from a bulletin board with a man on it to a wine glass with a yellowish tint, as well as the front direction of a bulletin board with a man on it. Next we compute the cosine similarities between the vector and the front direction, which leads to the angle between the vector and the fron direction. If the angle is smaller than 90 degrees, then a wine glass with a yellowish tint is in front of a bulletin board with a man on it. Otherwise, a wine glass with a yellowish tint is behind a bulletin board with a man on it. The 3D location of a wine glass with a yellowish tint is (-0.0, 0.2, 1.1). The 3D location of a bulletin board with a man on it is (0.6, 0.8, 1.5). The vector from a bulletin board with a man on it to a wine glass with a yellowish tint is hence (-0.7, -0.6, -0.4). The front direction of a bulletin board with a man on it is (-0.5, -0.1, -0.9). The cosine similarity between the vector and the front direction is 0.75, corresponding to an angle of 41.55 degrees. The angle is smaller than 90 degrees, meaning that a wine glass with a yellowish tint is in front of a bulletin board with a man on it. Therefore, the final answer is A. in front of.
|
A. in front of.
|
orientation_in_front_of
|
0049a8ab0b787a1a.jpg
|
0049ed52f4b63512_1fef
|
Consider the real-world 3D orientations of the objects. Are a black speaker with a white circle and a black speaker with a silver circle in the middle facing same or similar directions, or very different directions?
|
same or similar directions
|
very different directions
| null | null |
[
{
"bbox_3d": [
0.3,
2.4,
4.4
],
"label": "a black speaker with a white circle"
},
{
"bbox_3d": [
-0.4,
1.9,
3.9
],
"label": "a black speaker with a silver circle in the middle"
}
] |
[
{
"front_dir": [
0,
-0.4,
-0.9
],
"label": "a black speaker with a white circle",
"left_dir": [
-1,
0,
0
]
},
{
"front_dir": [
0.2,
-0.3,
-0.9
],
"label": "a black speaker with a silver circle in the middle",
"left_dir": [
-1,
0,
-0.2
]
}
] |
A
|
To solve this problem, we first detect the front directions of a black speaker with a white circle and a black speaker with a silver circle in the middle. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is small, then the two objects are facing same or similar directions. Otherwise, the two objects are facing very different directions. The front direction of a black speaker with a white circle is (0.0, -0.4, -0.9). The front direction of a black speaker with a silver circle in the middle is (0.2, -0.3, -0.9). The cosine similarity between the two front directions is 0.98, corresponding to an angle of 10.74. The angle is small, meaning that the two objects are facing same or similar directions. Therefore, the final answer is A. same or similar directions.
|
A. same or similar directions.
|
multi_object_same_direction
|
0049ed52f4b63512.jpg
|
004a140725bc9baf_1c0f
|
Consider the real-world 3D locations and orientations of the objects. Which side of a black motorcycle is facing a small blue car?
|
front
|
left
|
back
|
right
|
[
{
"bbox_3d": [
0.4,
1.1,
14.7
],
"label": "a black motorcycle"
},
{
"bbox_3d": [
4.2,
1.5,
17.5
],
"label": "a small blue car"
}
] |
[
{
"front_dir": [
0.6,
-0.2,
-0.8
],
"label": "a black motorcycle",
"left_dir": [
-0.8,
-0.1,
-0.6
]
},
{
"front_dir": [
-0.2,
-0.1,
-1
],
"label": "a small blue car",
"left_dir": [
-1,
0.1,
0.2
]
}
] |
D
|
To solve this problem, we first detect the 3D locations of a black motorcycle and a small blue car. Then we compute the vector pointing from a black motorcycle to a small blue car. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a black motorcycle, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a black motorcycle that is facing a small blue car corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The 3D location of a black motorcycle is (0.4, 1.1, 14.7). The 3D location of a small blue car is (4.2, 1.5, 17.5). The vector from a black motorcycle to a small blue car is hence (3.7, 0.4, 2.7). The left direction of a black motorcycle is (-0.8, -0.1, -0.6). The cosine similarity between the vector pointing to a small blue car and the left direction is -1.00, corresponding to an angle of 178.39 degrees. Thus the angle between the vector pointing to a small blue car and the right direction is 1.61 degrees. The front direction of a black motorcycle is (0.6, -0.2, -0.8). The cosine similarity between the vector pointing to a small blue car and the front direction is 0.03, corresponding to an angle of 88.39 degrees. Thus the angle between the vector pointing to a small blue car and the back direction is 91.61 degrees. Among the four directions, the smallest angle is the right direction, with an angle of 1.61 degrees. Thus the right side of a black motorcycle is facing the a small blue car. Therefore, the final answer is D. right.
|
D. right.
|
multi_object_viewpoint_towards_object
|
004a140725bc9baf.jpg
|
004acfb339e12aa6_f961
|
Consider the real-world 3D locations and orientations of the objects. Which side of a silver car with a sun roof is facing a car with a brown seat?
|
front
|
left
|
back
|
right
|
[
{
"bbox_3d": [
-2.4,
1.5,
9
],
"label": "a silver car with a sun roof"
},
{
"bbox_3d": [
1.1,
0.7,
1.7
],
"label": "a car with a brown seat"
}
] |
[
{
"front_dir": [
0.6,
0,
-0.8
],
"label": "a silver car with a sun roof",
"left_dir": [
-0.8,
-0.1,
-0.6
]
},
{
"front_dir": [
-1,
-0.1,
0.1
],
"label": "a car with a brown seat",
"left_dir": [
0.1,
-0.1,
1
]
}
] |
A
|
To solve this problem, we first detect the 3D locations of a silver car with a sun roof and a car with a brown seat. Then we compute the vector pointing from a silver car with a sun roof to a car with a brown seat. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a silver car with a sun roof, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a silver car with a sun roof that is facing a car with a brown seat corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The 3D location of a silver car with a sun roof is (-2.4, 1.5, 9.0). The 3D location of a car with a brown seat is (1.1, 0.7, 1.7). The vector from a silver car with a sun roof to a car with a brown seat is hence (3.5, -0.8, -7.3). The left direction of a silver car with a sun roof is (-0.8, -0.1, -0.6). The cosine similarity between the vector pointing to a car with a brown seat and the left direction is 0.22, corresponding to an angle of 77.52 degrees. Thus the angle between the vector pointing to a car with a brown seat and the right direction is 102.48 degrees. The front direction of a silver car with a sun roof is (0.6, -0.0, -0.8). The cosine similarity between the vector pointing to a car with a brown seat and the front direction is 0.98, corresponding to an angle of 12.73 degrees. Thus the angle between the vector pointing to a car with a brown seat and the back direction is 167.27 degrees. Among the four directions, the smallest angle is the front direction, with an angle of 12.73 degrees. Thus the front side of a silver car with a sun roof is facing the a car with a brown seat. Therefore, the final answer is A. front.
|
A. front.
|
multi_object_viewpoint_towards_object
|
004acfb339e12aa6.jpg
|
004ad904abc3534b_a652
|
Consider the real-world 3D locations of the objects. Are the a stone gazebo with a wooden roof and the a garden with a lot of plants next to each other or far away from each other?
|
next to each other
|
far away from each other
| null | null |
[
{
"bbox_3d": [
6,
3.7,
15.7
],
"label": "a stone gazebo with a wooden roof"
},
{
"bbox_3d": [
0.8,
0.6,
4.4
],
"label": "a garden with a lot of plants"
}
] |
[] |
A
|
To solve this problem, we first estimate the 3D locations of a stone gazebo with a wooden roof and a garden with a lot of plants. Then we can compute the L2 distance between the two objects. Next we estimate the rough sizes of the two objects. If the distance between the two objects is smaller or roughly the same as the object sizes, then the two objects are next to each other. Otherwise, the two objects are far away from each other. The 3D location of a stone gazebo with a wooden roof is (6.0, 3.7, 15.7). The 3D location of a garden with a lot of plants is (0.8, 0.6, 4.4). The L2 distance between the two objects is 12.79. The size of the a stone gazebo with a wooden roof is roughly 17.22. The size of the a garden with a lot of plants is roughly 5.43. The distance between the two objects is smaller or roughly the same as the object sizes, meaning that they are next to each other. Therefore, the answer is A. next to each other.
|
A. next to each other.
|
location_next_to
|
004ad904abc3534b.jpg
|
004b8cc59eaa3bbc_91ab
|
Consider the real-world 3D locations and orientations of the objects. Which object is a bookshelf with music cds facing towards, a bookshelf with white shelves or the a bookstore with aisles of books and electronics?
|
a bookshelf with white shelves
|
a bookstore with aisles of books and electronics
| null | null |
[
{
"bbox_3d": [
-0.5,
0.9,
2.4
],
"label": "a bookshelf with music cds"
},
{
"bbox_3d": [
1.4,
0.9,
2.3
],
"label": "a bookshelf with white shelves"
},
{
"bbox_3d": [
-0.8,
1.8,
2
],
"label": "a bookstore with aisles of books and electronics"
}
] |
[
{
"front_dir": [
1,
-0.1,
-0.1
],
"label": "a bookshelf with music cds",
"left_dir": [
0,
0.3,
-1
]
},
{
"front_dir": [
-0.6,
-0.1,
-0.8
],
"label": "a bookshelf with white shelves",
"left_dir": [
-0.8,
-0.1,
0.6
]
}
] |
A
|
To solve this problem, we first detect the 3D location of a bookshelf with music cds, a bookshelf with white shelves, and a bookstore with aisles of books and electronics. Then we compute the cosine similarities between the front direction of a bookshelf with music cds and the vectors from a bookshelf with music cds to the other two objects. We can estimate the angles from the cosine similarities, and the object with a smaller angle is the object that a bookshelf with music cds is facing towards. The 3D location of a bookshelf with music cds is (-0.5, 0.9, 2.4). The 3D location of a bookshelf with white shelves is (1.4, 0.9, 2.3). The 3D location of a bookstore with aisles of books and electronics is (-0.8, 1.8, 2.0). The front direction of a bookshelf with music cds is (1.0, -0.1, -0.1). First we consider if a bookshelf with music cds is facing towards the a bookshelf with white shelves. The vector from a bookshelf with music cds to a bookshelf with white shelves is (2.0, -0.0, -0.1). The cosine similarity between the front direction and the vector is 1.00, corresponding to an angle of 4.91 degrees. First we consider if a bookshelf with music cds is facing towards the a bookstore with aisles of books and electronics. The vector from a bookshelf with music cds to a bookstore with aisles of books and electronics is (-0.2, 0.8, -0.4). The cosine similarity between the front direction and the vector is -0.30, corresponding to an angle of 107.36 degrees. We find that the angle between the front direction and a bookshelf with white shelves is smaller. Therefore, the final answer is A. a bookshelf with white shelves.
|
A. a bookshelf with white shelves.
|
multi_object_facing
|
004b8cc59eaa3bbc.jpg
|
004bfdcce4cebc73_f5f1
|
Consider the real-world 3D locations of the objects. Is a surfer in the water directly underneath a green and yellow sail?
|
yes
|
no
| null | null |
[
{
"bbox_3d": [
-0.4,
2.8,
7.6
],
"label": "a green and yellow sail"
},
{
"bbox_3d": [
1.4,
0.9,
10.3
],
"label": "a surfer in the water"
}
] |
[] |
B
|
To solve this problem, we first determine the 3D locations of a green and yellow sail and a surfer in the water. Then we compute the vector pointing from a surfer in the water to a green and yellow sail, as well as the up direction of a surfer in the water. We estimate the cosine similarity between the vector and the up direction, which leads to the angle between the two directions. If the angle is small, then a green and yellow sail is directly above a surfer in the water. Otherwise, then a green and yellow sail is not directly above a surfer in the water. To solve the question, we first determine if a green and yellow sail is directly above a surfer in the water. The 3D location of a green and yellow sail is (-0.4, 2.8, 7.6). The 3D location of a surfer in the water is (1.4, 0.9, 10.3). The vector from a surfer in the water to a green and yellow sail is hence (-1.9, 1.9, -2.7). The up direction of a surfer in the water is (0.0, 1.0, 0.0). The cosine similarity between the vector and the up direction is 0.50, corresponding to an angle of 59 degrees. The angle between the vector and the up direction is large, meaning that a green and yellow sail is not directly above a surfer in the water. In other words, a surfer in the water is not directly underneath a green and yellow sail. Therefore, the answer is B. no.
|
B. no.
|
location_above
|
004bfdcce4cebc73.jpg
|
004c1c2b94414d04_14d4
|
Consider the real-world 3D orientations of the objects. What is the relationship between the orientations of a yellow car and a yellow car with a black stripe on the side, parallel of perpendicular to each other?
|
parallel
|
perpendicular
| null | null |
[
{
"bbox_3d": [
-2.7,
1.2,
6.4
],
"label": "a yellow car"
},
{
"bbox_3d": [
-0.2,
0.1,
3.4
],
"label": "a yellow car with a black stripe on the side"
}
] |
[
{
"front_dir": [
0.9,
0.1,
0.5
],
"label": "a yellow car",
"left_dir": [
0.5,
-0.1,
-0.9
]
},
{
"front_dir": [
1,
-0.1,
0.1
],
"label": "a yellow car with a black stripe on the side",
"left_dir": [
0.1,
0.4,
-0.9
]
}
] |
A
|
To solve this problem, we first detect the front directions of a yellow car and a yellow car with a black stripe on the side. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is closer to 0 or 180, than to 90 degrees, then the two objects are parallel to each other. Otherwise, the two objects are perpendicular to each other. The front direction of a yellow car is (0.9, 0.1, 0.5). The front direction of a yellow car with a black stripe on the side is (1.0, -0.1, 0.1). The cosine similarity between the two front directions is 0.92, corresponding to an angle of 23.30. The angle is closer to 0 or 180, than to 90 degrees, meaning that the two objects are parallel to each other. Therefore, the final answer is A. parallel.
|
A. parallel.
|
multi_object_parallel
|
004c1c2b94414d04.jpg
|
004ca3ed2ffdf4e8_6113
|
Consider the real-world 3D locations and orientations of the objects. If I stand at a red cart with a basket's position facing where it is facing, is a dark bottle with a label in front of me or behind me?
|
in front of
|
behind
| null | null |
[
{
"bbox_3d": [
-0.7,
0.5,
1.9
],
"label": "a dark bottle with a label"
},
{
"bbox_3d": [
0.1,
1,
0.9
],
"label": "a red cart with a basket"
}
] |
[
{
"front_dir": [
0,
0.4,
-0.9
],
"label": "a red cart with a basket",
"left_dir": [
-1,
0.1,
0.1
]
}
] |
B
|
To solve this problem, we first determine the 3D locations of a dark bottle with a label and a red cart with a basket. Then we estimate the vector pointing from a red cart with a basket to a dark bottle with a label, as well as the front direction of a red cart with a basket. Next we compute the cosine similarities between the vector and the front direction, which leads to the angle between the vector and the fron direction. If the angle is smaller than 90 degrees, then a dark bottle with a label is in front of a red cart with a basket. Otherwise, a dark bottle with a label is behind a red cart with a basket. The 3D location of a dark bottle with a label is (-0.7, 0.5, 1.9). The 3D location of a red cart with a basket is (0.1, 1.0, 0.9). The vector from a red cart with a basket to a dark bottle with a label is hence (-0.8, -0.5, 1.0). The front direction of a red cart with a basket is (-0.0, 0.4, -0.9). The cosine similarity between the vector and the front direction is -0.80, corresponding to an angle of 143.47 degrees. The angle is smaller than 90 degrees, meaning that a dark bottle with a label is behind a red cart with a basket. Therefore, the final answer is B. behind.
|
B. behind.
|
orientation_in_front_of
|
004ca3ed2ffdf4e8.jpg
|
004e43243709ea00_3558
|
Consider the real-world 3D locations of the objects. Which object has a lower location?
|
a man in a brown hat sitting
|
a dog with a white face and brown body
| null | null |
[
{
"bbox_3d": [
3.4,
0.5,
8.2
],
"label": "a man in a brown hat sitting"
},
{
"bbox_3d": [
1,
0.4,
4
],
"label": "a dog with a white face and brown body"
}
] |
[] |
B
|
To solve this problem, we first estimate the 3D heights of the two objects. The object with a larger height value is at a higher location. The object with a smaller height value is at a lower location. The 3D height of a man in a brown hat sitting is 1.6. The 3D height of a dog with a white face and brown body is 0.7. The 3D height of a man in a brown hat sitting is larger, meaning that the location of a man in a brown hat sitting is higher. In other words, the location of a dog with a white face and brown body is lower. Therefore, the answer is B. a man in a brown hat sitting.
|
B. a man in a brown hat sitting
|
height_higher
|
004e43243709ea00.jpg
|
004e43243709ea00_c48c
|
Consider the real-world 3D locations and orientations of the objects. Which side of a cart with a yellow cover is facing a white dog with a red leash?
|
front
|
left
|
back
|
right
|
[
{
"bbox_3d": [
3.4,
0.1,
16.5
],
"label": "a cart with a yellow cover"
},
{
"bbox_3d": [
-0.3,
0.4,
3.9
],
"label": "a white dog with a red leash"
}
] |
[
{
"front_dir": [
-0.1,
0,
-1
],
"label": "a cart with a yellow cover",
"left_dir": [
-1,
0,
0.1
]
}
] |
A
|
To solve this problem, we first detect the 3D locations of a cart with a yellow cover and a white dog with a red leash. Then we compute the vector pointing from a cart with a yellow cover to a white dog with a red leash. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a cart with a yellow cover, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a cart with a yellow cover that is facing a white dog with a red leash corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The 3D location of a cart with a yellow cover is (3.4, 0.1, 16.5). The 3D location of a white dog with a red leash is (-0.3, 0.4, 3.9). The vector from a cart with a yellow cover to a white dog with a red leash is hence (-3.7, 0.3, -12.5). The left direction of a cart with a yellow cover is (-1.0, 0.0, 0.1). The cosine similarity between the vector pointing to a white dog with a red leash and the left direction is 0.16, corresponding to an angle of 80.82 degrees. Thus the angle between the vector pointing to a white dog with a red leash and the right direction is 99.18 degrees. The front direction of a cart with a yellow cover is (-0.1, 0.0, -1.0). The cosine similarity between the vector pointing to a white dog with a red leash and the front direction is 0.99, corresponding to an angle of 9.18 degrees. Thus the angle between the vector pointing to a white dog with a red leash and the back direction is 170.82 degrees. Among the four directions, the smallest angle is the front direction, with an angle of 9.18 degrees. Thus the front side of a cart with a yellow cover is facing the a white dog with a red leash. Therefore, the final answer is A. front.
|
A. front.
|
multi_object_viewpoint_towards_object
|
004e43243709ea00.jpg
|
004e8f8aa9860565_8dd0
|
Consider the real-world 3D location of the objects. Which object is further away from the camera?
|
a red apple with a stem
|
a white bowl filled with cereal
| null | null |
[
{
"bbox_3d": [
0.2,
0.1,
0.2
],
"label": "a red apple with a stem"
},
{
"bbox_3d": [
-0.1,
0.1,
0.4
],
"label": "a white bowl filled with cereal"
}
] |
[] |
B
|
To solve this problem, we first estimate the 3D locations of a red apple with a stem and a white bowl filled with cereal. Then we estimate the L2 distances from the camera to the two objects. The object with a smaller L2 distance is the one that is closer to the camera. The 3D location of a red apple with a stem is (0.2, 0.1, 0.2). The 3D location of a white bowl filled with cereal is (-0.1, 0.1, 0.4). The L2 distance from the camera to a red apple with a stem is 0.30. The L2 distance from the camera to a white bowl filled with cereal is 0.44. The distance to a white bowl filled with cereal is larger. Therefore, the answer is B. a white bowl filled with cereal.
|
B. a white bowl filled with cereal.
|
location_closer_to_camera
|
004e8f8aa9860565.jpg
|
004f45f657786da9_0c67
|
Consider the real-world 3D location of the objects. Which object is closer to the camera?
|
a man wearing a white shirt and a blue tie
|
a red tie
| null | null |
[
{
"bbox_3d": [
1.2,
0.8,
3.4
],
"label": "a man wearing a white shirt and a blue tie"
},
{
"bbox_3d": [
-0.2,
1.3,
1.7
],
"label": "a red tie"
}
] |
[] |
B
|
To solve this problem, we first estimate the 3D locations of a man wearing a white shirt and a blue tie and a red tie. Then we estimate the L2 distances from the camera to the two objects. The object with a smaller L2 distance is the one that is closer to the camera. The 3D location of a man wearing a white shirt and a blue tie is (1.2, 0.8, 3.4). The 3D location of a red tie is (-0.2, 1.3, 1.7). The L2 distance from the camera to a man wearing a white shirt and a blue tie is 3.71. The L2 distance from the camera to a red tie is 2.09. The distance to a red tie is smaller. Therefore, the answer is B. a red tie.
|
B. a red tie.
|
location_closer_to_camera
|
004f45f657786da9.jpg
|
004f7aac33f4f2f6_83af
|
Consider the real-world 3D locations and orientations of the objects. If I stand at a white computer monitor's position facing where it is facing, is a wood wall with a red design in front of me or behind me?
|
in front of
|
behind
| null | null |
[
{
"bbox_3d": [
-0.1,
1.7,
4
],
"label": "a wood wall with a red design"
},
{
"bbox_3d": [
-0.1,
1.5,
2.1
],
"label": "a white computer monitor"
}
] |
[
{
"front_dir": [
-0.3,
-0.2,
-0.9
],
"label": "a white computer monitor",
"left_dir": [
-0.9,
0.1,
0.3
]
}
] |
B
|
To solve this problem, we first determine the 3D locations of a wood wall with a red design and a white computer monitor. Then we estimate the vector pointing from a white computer monitor to a wood wall with a red design, as well as the front direction of a white computer monitor. Next we compute the cosine similarities between the vector and the front direction, which leads to the angle between the vector and the fron direction. If the angle is smaller than 90 degrees, then a wood wall with a red design is in front of a white computer monitor. Otherwise, a wood wall with a red design is behind a white computer monitor. The 3D location of a wood wall with a red design is (-0.1, 1.7, 4.0). The 3D location of a white computer monitor is (-0.1, 1.5, 2.1). The vector from a white computer monitor to a wood wall with a red design is hence (-0.1, 0.2, 1.9). The front direction of a white computer monitor is (-0.3, -0.2, -0.9). The cosine similarity between the vector and the front direction is -0.93, corresponding to an angle of 157.92 degrees. The angle is smaller than 90 degrees, meaning that a wood wall with a red design is behind a white computer monitor. Therefore, the final answer is B. behind.
|
B. behind.
|
orientation_in_front_of
|
004f7aac33f4f2f6.jpg
|
00502e0b65d0eee0_a0ea
|
Consider the real-world 3D locations of the objects. Are the a black stool and the a person in a wheelchair with a red shirt next to each other or far away from each other?
|
next to each other
|
far away from each other
| null | null |
[
{
"bbox_3d": [
1.2,
0.4,
2.7
],
"label": "a black stool"
},
{
"bbox_3d": [
-0.3,
0.8,
3.2
],
"label": "a person in a wheelchair with a red shirt"
}
] |
[] |
A
|
To solve this problem, we first estimate the 3D locations of a black stool and a person in a wheelchair with a red shirt. Then we can compute the L2 distance between the two objects. Next we estimate the rough sizes of the two objects. If the distance between the two objects is smaller or roughly the same as the object sizes, then the two objects are next to each other. Otherwise, the two objects are far away from each other. The 3D location of a black stool is (1.2, 0.4, 2.7). The 3D location of a person in a wheelchair with a red shirt is (-0.3, 0.8, 3.2). The L2 distance between the two objects is 1.66. The size of the a black stool is roughly 0.82. The size of the a person in a wheelchair with a red shirt is roughly 1.42. The distance between the two objects is smaller or roughly the same as the object sizes, meaning that they are next to each other. Therefore, the answer is A. next to each other.
|
A. next to each other.
|
location_next_to
|
00502e0b65d0eee0.jpg
|
00519c12e294814b_d507
|
Consider the real-world 3D locations and orientations of the objects. Which side of a wooden handcart with bananas is facing a boat with the word "ota" on it?
|
front
|
left
|
back
|
right
|
[
{
"bbox_3d": [
0.3,
0.5,
5.6
],
"label": "a wooden handcart with bananas"
},
{
"bbox_3d": [
-7.3,
-0.2,
28.6
],
"label": "a boat with the word \"ota\" on it"
}
] |
[
{
"front_dir": [
0,
0,
-1
],
"label": "a wooden handcart with bananas",
"left_dir": [
-1,
0.1,
0
]
}
] |
C
|
To solve this problem, we first detect the 3D locations of a wooden handcart with bananas and a boat with the word "ota" on it. Then we compute the vector pointing from a wooden handcart with bananas to a boat with the word "ota" on it. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a wooden handcart with bananas, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a wooden handcart with bananas that is facing a boat with the word "ota" on it corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The 3D location of a wooden handcart with bananas is (0.3, 0.5, 5.6). The 3D location of a boat with the word "ota" on it is (-7.3, -0.2, 28.6). The vector from a wooden handcart with bananas to a boat with the word "ota" on it is hence (-7.5, -0.7, 23.1). The left direction of a wooden handcart with bananas is (-1.0, 0.1, -0.0). The cosine similarity between the vector pointing to a boat with the word "ota" on it and the left direction is 0.29, corresponding to an angle of 73.32 degrees. Thus the angle between the vector pointing to a boat with the word "ota" on it and the right direction is 106.68 degrees. The front direction of a wooden handcart with bananas is (0.0, -0.0, -1.0). The cosine similarity between the vector pointing to a boat with the word "ota" on it and the front direction is -0.95, corresponding to an angle of 162.68 degrees. Thus the angle between the vector pointing to a boat with the word "ota" on it and the back direction is 17.32 degrees. Among the four directions, the smallest angle is the back direction, with an angle of 17.32 degrees. Thus the back side of a wooden handcart with bananas is facing the a boat with the word "ota" on it. Therefore, the final answer is C. back.
|
C. back.
|
multi_object_viewpoint_towards_object
|
00519c12e294814b.jpg
|
005220abf30b472f_18f2
|
Consider the real-world 3D locations and orientations of the objects. If I stand at a brown leather armchair's position facing where it is facing, is a white vase with a pattern in front of me or behind me?
|
in front of
|
behind
| null | null |
[
{
"bbox_3d": [
-1.4,
0.4,
2.2
],
"label": "a white vase with a pattern"
},
{
"bbox_3d": [
1.3,
0.3,
5.2
],
"label": "a brown leather armchair"
}
] |
[
{
"front_dir": [
-0.7,
0,
-0.7
],
"label": "a brown leather armchair",
"left_dir": [
-0.7,
0.1,
0.7
]
}
] |
A
|
To solve this problem, we first determine the 3D locations of a white vase with a pattern and a brown leather armchair. Then we estimate the vector pointing from a brown leather armchair to a white vase with a pattern, as well as the front direction of a brown leather armchair. Next we compute the cosine similarities between the vector and the front direction, which leads to the angle between the vector and the fron direction. If the angle is smaller than 90 degrees, then a white vase with a pattern is in front of a brown leather armchair. Otherwise, a white vase with a pattern is behind a brown leather armchair. The 3D location of a white vase with a pattern is (-1.4, 0.4, 2.2). The 3D location of a brown leather armchair is (1.3, 0.3, 5.2). The vector from a brown leather armchair to a white vase with a pattern is hence (-2.8, 0.1, -3.0). The front direction of a brown leather armchair is (-0.7, -0.0, -0.7). The cosine similarity between the vector and the front direction is 1.00, corresponding to an angle of 5.49 degrees. The angle is smaller than 90 degrees, meaning that a white vase with a pattern is in front of a brown leather armchair. Therefore, the final answer is A. in front of.
|
A. in front of.
|
orientation_in_front_of
|
005220abf30b472f.jpg
|
0052c162aa0e8c22_d789
|
Consider the real-world 3D locations of the objects. Are the a man in a military uniform and the a man in a military uniform next to each other or far away from each other?
|
next to each other
|
far away from each other
| null | null |
[
{
"bbox_3d": [
0.1,
0.9,
3.2
],
"label": "a man in a military uniform"
},
{
"bbox_3d": [
-0.3,
0.6,
2.7
],
"label": "a man in a military uniform"
}
] |
[] |
A
|
To solve this problem, we first estimate the 3D locations of a man in a military uniform and a man in a military uniform. Then we can compute the L2 distance between the two objects. Next we estimate the rough sizes of the two objects. If the distance between the two objects is smaller or roughly the same as the object sizes, then the two objects are next to each other. Otherwise, the two objects are far away from each other. The 3D location of a man in a military uniform is (0.1, 0.9, 3.2). The 3D location of a man in a military uniform is (-0.3, 0.6, 2.7). The L2 distance between the two objects is 0.74. The size of the a man in a military uniform is roughly 1.37. The size of the a man in a military uniform is roughly 1.20. The distance between the two objects is smaller or roughly the same as the object sizes, meaning that they are next to each other. Therefore, the answer is A. next to each other.
|
A. next to each other.
|
location_next_to
|
0052c162aa0e8c22.jpg
|
0053704554a2ea33_53b1
|
Consider the real-world 3D locations and orientations of the objects. Which side of a blue and white space shuttle is facing a blue balloon?
|
front
|
left
|
back
|
right
|
[
{
"bbox_3d": [
0.3,
1.6,
1.7
],
"label": "a blue and white space shuttle"
},
{
"bbox_3d": [
-0.1,
1.5,
1.5
],
"label": "a blue balloon"
}
] |
[
{
"front_dir": [
0.4,
-0.3,
-0.9
],
"label": "a blue and white space shuttle",
"left_dir": [
-0.9,
-0.3,
-0.3
]
}
] |
B
|
To solve this problem, we first detect the 3D locations of a blue and white space shuttle and a blue balloon. Then we compute the vector pointing from a blue and white space shuttle to a blue balloon. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a blue and white space shuttle, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a blue and white space shuttle that is facing a blue balloon corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The 3D location of a blue and white space shuttle is (0.3, 1.6, 1.7). The 3D location of a blue balloon is (-0.1, 1.5, 1.5). The vector from a blue and white space shuttle to a blue balloon is hence (-0.4, -0.1, -0.2). The left direction of a blue and white space shuttle is (-0.9, -0.3, -0.3). The cosine similarity between the vector pointing to a blue balloon and the left direction is 0.98, corresponding to an angle of 12.07 degrees. Thus the angle between the vector pointing to a blue balloon and the right direction is 167.93 degrees. The front direction of a blue and white space shuttle is (0.4, -0.3, -0.9). The cosine similarity between the vector pointing to a blue balloon and the front direction is 0.08, corresponding to an angle of 85.21 degrees. Thus the angle between the vector pointing to a blue balloon and the back direction is 94.79 degrees. Among the four directions, the smallest angle is the left direction, with an angle of 12.07 degrees. Thus the left side of a blue and white space shuttle is facing the a blue balloon. Therefore, the final answer is B. left.
|
B. left.
|
multi_object_viewpoint_towards_object
|
0053704554a2ea33.jpg
|
0053939eb128715f_6b0c
|
Consider the real-world 3D locations of the objects. Which object has a lower location?
|
a dock with a large boat
|
a large white boat
| null | null |
[
{
"bbox_3d": [
-10.3,
8.4,
107.9
],
"label": "a dock with a large boat"
},
{
"bbox_3d": [
20,
16.3,
156.9
],
"label": "a large white boat"
}
] |
[] |
A
|
To solve this problem, we first estimate the 3D heights of the two objects. The object with a larger height value is at a higher location. The object with a smaller height value is at a lower location. The 3D height of a dock with a large boat is 26.7. The 3D height of a large white boat is 72.4. The 3D height of a large white boat is larger, meaning that the location of a large white boat is higher. In other words, the location of a dock with a large boat is lower. Therefore, the answer is A. a dock with a large boat.
|
A. a dock with a large boat
|
height_higher
|
0053939eb128715f.jpg
|
0054705f13ab32a0_2c3f
|
Consider the real-world 3D location of the objects. Which object is closer to the camera?
|
a large white boat in the water
|
a helicopter with a rope
| null | null |
[
{
"bbox_3d": [
-4.1,
1.6,
61.8
],
"label": "a large white boat in the water"
},
{
"bbox_3d": [
-24.7,
19.1,
132.2
],
"label": "a helicopter with a rope"
}
] |
[] |
A
|
To solve this problem, we first estimate the 3D locations of a large white boat in the water and a helicopter with a rope. Then we estimate the L2 distances from the camera to the two objects. The object with a smaller L2 distance is the one that is closer to the camera. The 3D location of a large white boat in the water is (-4.1, 1.6, 61.8). The 3D location of a helicopter with a rope is (-24.7, 19.1, 132.2). The L2 distance from the camera to a large white boat in the water is 61.93. The L2 distance from the camera to a helicopter with a rope is 135.79. The distance to a large white boat in the water is smaller. Therefore, the answer is A. a large white boat in the water.
|
A. a large white boat in the water.
|
location_closer_to_camera
|
0054705f13ab32a0.jpg
|
005553035e586a56_d8b6
|
Consider the real-world 3D locations and orientations of the objects. Which object is a boat with people on it facing towards, a body of water or the a metal rail?
|
a body of water
|
a metal rail
| null | null |
[
{
"bbox_3d": [
1.9,
1.7,
17.8
],
"label": "a boat with people on it"
},
{
"bbox_3d": [
-2.8,
0.9,
18
],
"label": "a body of water"
},
{
"bbox_3d": [
-0.6,
1.1,
1.7
],
"label": "a metal rail"
}
] |
[
{
"front_dir": [
0,
-0.5,
-0.9
],
"label": "a boat with people on it",
"left_dir": [
-1,
0,
0
]
}
] |
B
|
To solve this problem, we first detect the 3D location of a boat with people on it, a body of water, and a metal rail. Then we compute the cosine similarities between the front direction of a boat with people on it and the vectors from a boat with people on it to the other two objects. We can estimate the angles from the cosine similarities, and the object with a smaller angle is the object that a boat with people on it is facing towards. The 3D location of a boat with people on it is (1.9, 1.7, 17.8). The 3D location of a body of water is (-2.8, 0.9, 18.0). The 3D location of a metal rail is (-0.6, 1.1, 1.7). The front direction of a boat with people on it is (-0.0, -0.5, -0.9). First we consider if a boat with people on it is facing towards the a body of water. The vector from a boat with people on it to a body of water is (-4.7, -0.8, 0.2). The cosine similarity between the front direction and the vector is 0.06, corresponding to an angle of 86.37 degrees. First we consider if a boat with people on it is facing towards the a metal rail. The vector from a boat with people on it to a metal rail is (-2.4, -0.6, -16.1). The cosine similarity between the front direction and the vector is 0.88, corresponding to an angle of 28.66 degrees. We find that the angle between the front direction and a metal rail is smaller. Therefore, the final answer is B. a metal rail.
|
B. a metal rail.
|
multi_object_facing
|
005553035e586a56.jpg
|
0056b888593cdc08_dcfc
|
Consider the real-world 3D locations and orientations of the objects. If I stand at a brick building with a metal cannon's position facing where it is facing, is a tall metal tower with a dome on the left or right of me?
|
on the left
|
on the right
| null | null |
[
{
"bbox_3d": [
22.2,
9.7,
98.3
],
"label": "a tall metal tower with a dome"
},
{
"bbox_3d": [
15.1,
2.3,
54.8
],
"label": "a brick building with a metal cannon"
}
] |
[
{
"front_dir": [
-0.7,
0,
-0.7
],
"label": "a tall metal tower with a dome",
"left_dir": [
-0.7,
0.1,
0.7
]
},
{
"front_dir": [
-1,
0,
-0.3
],
"label": "a brick building with a metal cannon",
"left_dir": [
-0.3,
0.1,
1
]
}
] |
A
|
To solve this problem, we first determine the 3D locations of a tall metal tower with a dome and a brick building with a metal cannon. Then we estimate the vector pointing from a brick building with a metal cannon to a tall metal tower with a dome, as well as the left direction of a brick building with a metal cannon. Next we compute the cosine similarities between the vector and the left direction, which leads to the angle between the vector and the fron direction. If the angle is smaller than 90 degrees, then a tall metal tower with a dome is on the left of a brick building with a metal cannon. Otherwise, a tall metal tower with a dome is behind a brick building with a metal cannon. The 3D location of a tall metal tower with a dome is (22.2, 9.7, 98.3). The 3D location of a brick building with a metal cannon is (15.1, 2.3, 54.8). The vector from a brick building with a metal cannon to a tall metal tower with a dome is hence (7.1, 7.4, 43.5). The left direction of a brick building with a metal cannon is (-0.3, 0.1, 1.0). The cosine similarity between the vector and the left direction is 0.90, corresponding to an angle of 25.29 degrees. The angle is smaller than 90 degrees, meaning that a tall metal tower with a dome is on the left of a brick building with a metal cannon. Therefore, the final answer is A. on the left.
|
A. on the left.
|
orientation_on_the_left
|
0056b888593cdc08.jpg
|
0056ed0a304a6d82_c5c5
|
Consider the real-world 3D locations of the objects. Which object has a lower location?
|
a boy wearing a red hat
|
a person wearing a brown jacket
| null | null |
[
{
"bbox_3d": [
0.4,
4.9,
6.6
],
"label": "a boy wearing a red hat"
},
{
"bbox_3d": [
-0.8,
0.6,
22.8
],
"label": "a person wearing a brown jacket"
}
] |
[] |
B
|
To solve this problem, we first estimate the 3D heights of the two objects. The object with a larger height value is at a higher location. The object with a smaller height value is at a lower location. The 3D height of a boy wearing a red hat is 5.8. The 3D height of a person wearing a brown jacket is 0.8. The 3D height of a boy wearing a red hat is larger, meaning that the location of a boy wearing a red hat is higher. In other words, the location of a person wearing a brown jacket is lower. Therefore, the answer is B. a boy wearing a red hat.
|
B. a boy wearing a red hat
|
height_higher
|
0056ed0a304a6d82.jpg
|
0057375432ee179a_2c1d
|
Consider the real-world 3D orientations of the objects. Are a wooden stool with a dark brown top and a black speaker facing same or similar directions, or very different directions?
|
same or similar directions
|
very different directions
| null | null |
[
{
"bbox_3d": [
1,
0.7,
2.7
],
"label": "a wooden stool with a dark brown top"
},
{
"bbox_3d": [
-1.3,
1.7,
4.2
],
"label": "a black speaker"
}
] |
[
{
"front_dir": [
0.6,
0.1,
-0.8
],
"label": "a wooden stool with a dark brown top",
"left_dir": [
-0.8,
0,
-0.6
]
},
{
"front_dir": [
0.4,
-0.1,
-0.9
],
"label": "a black speaker",
"left_dir": [
-0.9,
0,
-0.4
]
}
] |
A
|
To solve this problem, we first detect the front directions of a wooden stool with a dark brown top and a black speaker. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is small, then the two objects are facing same or similar directions. Otherwise, the two objects are facing very different directions. The front direction of a wooden stool with a dark brown top is (0.6, 0.1, -0.8). The front direction of a black speaker is (0.4, -0.1, -0.9). The cosine similarity between the two front directions is 0.94, corresponding to an angle of 20.03. The angle is small, meaning that the two objects are facing same or similar directions. Therefore, the final answer is A. same or similar directions.
|
A. same or similar directions.
|
multi_object_same_direction
|
0057375432ee179a.jpg
|
00577a50532b6e9c_ba25
|
Consider the real-world 3D locations and orientations of the objects. Which side of a black car is facing a man in a blue shirt and gray pants?
|
front
|
left
|
back
|
right
|
[
{
"bbox_3d": [
-4.6,
0.9,
7.4
],
"label": "a black car"
},
{
"bbox_3d": [
-1.4,
0.9,
3.7
],
"label": "a man in a blue shirt and gray pants"
}
] |
[
{
"front_dir": [
0.6,
0,
-0.8
],
"label": "a black car",
"left_dir": [
-0.8,
0.2,
-0.6
]
}
] |
A
|
To solve this problem, we first detect the 3D locations of a black car and a man in a blue shirt and gray pants. Then we compute the vector pointing from a black car to a man in a blue shirt and gray pants. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a black car, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a black car that is facing a man in a blue shirt and gray pants corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The 3D location of a black car is (-4.6, 0.9, 7.4). The 3D location of a man in a blue shirt and gray pants is (-1.4, 0.9, 3.7). The vector from a black car to a man in a blue shirt and gray pants is hence (3.1, -0.0, -3.7). The left direction of a black car is (-0.8, 0.2, -0.6). The cosine similarity between the vector pointing to a man in a blue shirt and gray pants and the left direction is -0.08, corresponding to an angle of 94.35 degrees. Thus the angle between the vector pointing to a man in a blue shirt and gray pants and the right direction is 85.65 degrees. The front direction of a black car is (0.6, -0.0, -0.8). The cosine similarity between the vector pointing to a man in a blue shirt and gray pants and the front direction is 1.00, corresponding to an angle of 4.49 degrees. Thus the angle between the vector pointing to a man in a blue shirt and gray pants and the back direction is 175.51 degrees. Among the four directions, the smallest angle is the front direction, with an angle of 4.49 degrees. Thus the front side of a black car is facing the a man in a blue shirt and gray pants. Therefore, the final answer is A. front.
|
A. front.
|
multi_object_viewpoint_towards_object
|
00577a50532b6e9c.jpg
|
00577a50532b6e9c_39a2
|
Consider the real-world 3D locations and orientations of the objects. Which object is a black car facing towards, a man in a suit or the a man with white hair?
|
a man in a suit
|
a man with white hair
| null | null |
[
{
"bbox_3d": [
-4.6,
0.9,
7.4
],
"label": "a black car"
},
{
"bbox_3d": [
-1.6,
0.9,
3.1
],
"label": "a man in a suit"
},
{
"bbox_3d": [
-15.1,
1.8,
27.7
],
"label": "a man with white hair"
}
] |
[
{
"front_dir": [
0.6,
0,
-0.8
],
"label": "a black car",
"left_dir": [
-0.8,
0.2,
-0.6
]
}
] |
A
|
To solve this problem, we first detect the 3D location of a black car, a man in a suit, and a man with white hair. Then we compute the cosine similarities between the front direction of a black car and the vectors from a black car to the other two objects. We can estimate the angles from the cosine similarities, and the object with a smaller angle is the object that a black car is facing towards. The 3D location of a black car is (-4.6, 0.9, 7.4). The 3D location of a man in a suit is (-1.6, 0.9, 3.1). The 3D location of a man with white hair is (-15.1, 1.8, 27.7). The front direction of a black car is (0.6, -0.0, -0.8). First we consider if a black car is facing towards the a man in a suit. The vector from a black car to a man in a suit is (3.0, -0.0, -4.3). The cosine similarity between the front direction and the vector is 1.00, corresponding to an angle of 0.62 degrees. First we consider if a black car is facing towards the a man with white hair. The vector from a black car to a man with white hair is (-10.5, 1.0, 20.3). The cosine similarity between the front direction and the vector is -0.99, corresponding to an angle of 171.74 degrees. We find that the angle between the front direction and a man in a suit is smaller. Therefore, the final answer is A. a man in a suit.
|
A. a man in a suit.
|
multi_object_facing
|
00577a50532b6e9c.jpg
|
00578d46bc49510a_4f15
|
Consider the real-world 3D location of the objects. Which object is further away from the camera?
|
a man with a microphone
|
a man in a white shirt with his arms raised
| null | null |
[
{
"bbox_3d": [
1.7,
2.3,
7.4
],
"label": "a man with a microphone"
},
{
"bbox_3d": [
0,
3.1,
3.9
],
"label": "a man in a white shirt with his arms raised"
}
] |
[] |
A
|
To solve this problem, we first estimate the 3D locations of a man with a microphone and a man in a white shirt with his arms raised. Then we estimate the L2 distances from the camera to the two objects. The object with a smaller L2 distance is the one that is closer to the camera. The 3D location of a man with a microphone is (1.7, 2.3, 7.4). The 3D location of a man in a white shirt with his arms raised is (0.0, 3.1, 3.9). The L2 distance from the camera to a man with a microphone is 7.97. The L2 distance from the camera to a man in a white shirt with his arms raised is 4.98. The distance to a man with a microphone is larger. Therefore, the answer is A. a man with a microphone.
|
A. a man with a microphone.
|
location_closer_to_camera
|
00578d46bc49510a.jpg
|
00579edc6b1090d7_c594
|
Consider the real-world 3D locations and orientations of the objects. Which side of a bicycle with a black frame is facing the camera?
|
front
|
left
|
back
|
right
|
[
{
"bbox_3d": [
0.2,
1.4,
14
],
"label": "a bicycle with a black frame"
}
] |
[
{
"front_dir": [
1,
-0.1,
0.1
],
"label": "a bicycle with a black frame",
"left_dir": [
0.1,
0,
-1
]
}
] |
B
|
To solve this problem, we first estimate the 3D location of a bicycle with a black frame. Then we obtain the vector pointing from the object to the camera. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a bicycle with a black frame, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a bicycle with a black frame that is facing the camera corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The location of a bicycle with a black frame is (0.2, 1.4, 14.0). The vector from a bicycle with a black frame to camera is hence (-0.2, -1.4, -14.0). The left direction of a bicycle with a black frame is (0.1, 0.0, -1.0). The cosine similarity between the vector pointing to camera and the left direction is 0.99, corresponding to an angle of 8.49 degrees. Thus the angle between the vector pointing to camera and the right direction is 171.51 degrees. The front direction of a bicycle with a black frame is (1.0, -0.1, 0.1). The cosine similarity between the vector pointing to camera and the front direction is -0.06, corresponding to an angle of 93.42 degrees. Thus the angle between the vector pointing to camera and the back direction is 86.58 degrees. Among the four directions, the smallest angle is the left direction, with an angle of 8.49 degrees. Thus the left side of a bicycle with a black frame is facing the camera. Therefore, the final answer is B. left.
|
B. left.
|
orientation_viewpoint
|
00579edc6b1090d7.jpg
|
00579edc6b1090d7_769d
|
Consider the real-world 3D locations and orientations of the objects. Which side of a bicycle with a black handlebar is facing the camera?
|
front
|
left
|
back
|
right
|
[
{
"bbox_3d": [
-1.1,
2.4,
13.1
],
"label": "a bicycle with a black handlebar"
}
] |
[
{
"front_dir": [
1,
-0.1,
0.2
],
"label": "a bicycle with a black handlebar",
"left_dir": [
0.2,
0,
-1
]
}
] |
B
|
To solve this problem, we first estimate the 3D location of a bicycle with a black handlebar. Then we obtain the vector pointing from the object to the camera. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a bicycle with a black handlebar, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a bicycle with a black handlebar that is facing the camera corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The location of a bicycle with a black handlebar is (-1.1, 2.4, 13.1). The vector from a bicycle with a black handlebar to camera is hence (1.1, -2.4, -13.1). The left direction of a bicycle with a black handlebar is (0.2, -0.0, -1.0). The cosine similarity between the vector pointing to camera and the left direction is 0.99, corresponding to an angle of 8.97 degrees. Thus the angle between the vector pointing to camera and the right direction is 171.03 degrees. The front direction of a bicycle with a black handlebar is (1.0, -0.1, 0.2). The cosine similarity between the vector pointing to camera and the front direction is -0.06, corresponding to an angle of 93.31 degrees. Thus the angle between the vector pointing to camera and the back direction is 86.69 degrees. Among the four directions, the smallest angle is the left direction, with an angle of 8.97 degrees. Thus the left side of a bicycle with a black handlebar is facing the camera. Therefore, the final answer is B. left.
|
B. left.
|
orientation_viewpoint
|
00579edc6b1090d7.jpg
|
0059ff2d54a9769e_ea49
|
Consider the real-world 3D orientations of the objects. What is the relationship between the orientations of a propeller on a plane and a white and black biplane, parallel of perpendicular to each other?
|
parallel
|
perpendicular
| null | null |
[
{
"bbox_3d": [
1.3,
2,
29.1
],
"label": "a propeller on a plane"
},
{
"bbox_3d": [
-0.1,
1.7,
27.9
],
"label": "a white and black biplane"
}
] |
[
{
"front_dir": [
0.5,
0,
-0.9
],
"label": "a propeller on a plane",
"left_dir": [
-0.9,
0.1,
-0.5
]
},
{
"front_dir": [
0.5,
0.1,
-0.8
],
"label": "a white and black biplane",
"left_dir": [
-0.8,
-0.1,
-0.5
]
}
] |
A
|
To solve this problem, we first detect the front directions of a propeller on a plane and a white and black biplane. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is closer to 0 or 180, than to 90 degrees, then the two objects are parallel to each other. Otherwise, the two objects are perpendicular to each other. The front direction of a propeller on a plane is (0.5, 0.0, -0.9). The front direction of a white and black biplane is (0.5, 0.1, -0.8). The cosine similarity between the two front directions is 1.00, corresponding to an angle of 5.60. The angle is closer to 0 or 180, than to 90 degrees, meaning that the two objects are parallel to each other. Therefore, the final answer is A. parallel.
|
A. parallel.
|
multi_object_parallel
|
0059ff2d54a9769e.jpg
|
005b12889dc6b131_92eb
|
Consider the real-world 3D locations of the objects. Are the a carpeted floor and the a white radiator next to each other or far away from each other?
|
next to each other
|
far away from each other
| null | null |
[
{
"bbox_3d": [
0.1,
0.1,
3.2
],
"label": "a carpeted floor"
},
{
"bbox_3d": [
-0.3,
0.5,
5.2
],
"label": "a white radiator"
}
] |
[] |
A
|
To solve this problem, we first estimate the 3D locations of a carpeted floor and a white radiator. Then we can compute the L2 distance between the two objects. Next we estimate the rough sizes of the two objects. If the distance between the two objects is smaller or roughly the same as the object sizes, then the two objects are next to each other. Otherwise, the two objects are far away from each other. The 3D location of a carpeted floor is (0.1, 0.1, 3.2). The 3D location of a white radiator is (-0.3, 0.5, 5.2). The L2 distance between the two objects is 2.06. The size of the a carpeted floor is roughly 2.70. The size of the a white radiator is roughly 1.41. The distance between the two objects is smaller or roughly the same as the object sizes, meaning that they are next to each other. Therefore, the answer is A. next to each other.
|
A. next to each other.
|
location_next_to
|
005b12889dc6b131.jpg
|
005b64d7a7dd7db0_fe29
|
Consider the real-world 3D locations and orientations of the objects. Which side of a white boat with a canopy is facing a person in a blue shirt sitting on a chair?
|
front
|
left
|
back
|
right
|
[
{
"bbox_3d": [
1,
1.6,
29.5
],
"label": "a white boat with a canopy"
},
{
"bbox_3d": [
-13.9,
4.5,
77.2
],
"label": "a person in a blue shirt sitting on a chair"
}
] |
[
{
"front_dir": [
0.6,
-0.1,
-0.8
],
"label": "a white boat with a canopy",
"left_dir": [
-0.8,
0,
-0.6
]
}
] |
C
|
To solve this problem, we first detect the 3D locations of a white boat with a canopy and a person in a blue shirt sitting on a chair. Then we compute the vector pointing from a white boat with a canopy to a person in a blue shirt sitting on a chair. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a white boat with a canopy, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a white boat with a canopy that is facing a person in a blue shirt sitting on a chair corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The 3D location of a white boat with a canopy is (1.0, 1.6, 29.5). The 3D location of a person in a blue shirt sitting on a chair is (-13.9, 4.5, 77.2). The vector from a white boat with a canopy to a person in a blue shirt sitting on a chair is hence (-14.9, 2.9, 47.7). The left direction of a white boat with a canopy is (-0.8, 0.0, -0.6). The cosine similarity between the vector pointing to a person in a blue shirt sitting on a chair and the left direction is -0.36, corresponding to an angle of 110.92 degrees. Thus the angle between the vector pointing to a person in a blue shirt sitting on a chair and the right direction is 69.08 degrees. The front direction of a white boat with a canopy is (0.6, -0.1, -0.8). The cosine similarity between the vector pointing to a person in a blue shirt sitting on a chair and the front direction is -0.93, corresponding to an angle of 159.05 degrees. Thus the angle between the vector pointing to a person in a blue shirt sitting on a chair and the back direction is 20.95 degrees. Among the four directions, the smallest angle is the back direction, with an angle of 20.95 degrees. Thus the back side of a white boat with a canopy is facing the a person in a blue shirt sitting on a chair. Therefore, the final answer is C. back.
|
C. back.
|
multi_object_viewpoint_towards_object
|
005b64d7a7dd7db0.jpg
|
005ba732f68b2949_4799
|
Consider the real-world 3D locations and orientations of the objects. Which side of a wooden chair with a person sitting on it is facing the camera?
|
front
|
left
|
back
|
right
|
[
{
"bbox_3d": [
-0.1,
0.4,
3.2
],
"label": "a wooden chair with a person sitting on it"
}
] |
[
{
"front_dir": [
0.1,
0.2,
-1
],
"label": "a wooden chair with a person sitting on it",
"left_dir": [
-1,
0.3,
0
]
}
] |
A
|
To solve this problem, we first estimate the 3D location of a wooden chair with a person sitting on it. Then we obtain the vector pointing from the object to the camera. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a wooden chair with a person sitting on it, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a wooden chair with a person sitting on it that is facing the camera corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The location of a wooden chair with a person sitting on it is (-0.1, 0.4, 3.2). The vector from a wooden chair with a person sitting on it to camera is hence (0.1, -0.4, -3.2). The left direction of a wooden chair with a person sitting on it is (-1.0, 0.3, -0.0). The cosine similarity between the vector pointing to camera and the left direction is -0.04, corresponding to an angle of 92.50 degrees. Thus the angle between the vector pointing to camera and the right direction is 87.50 degrees. The front direction of a wooden chair with a person sitting on it is (0.1, 0.2, -1.0). The cosine similarity between the vector pointing to camera and the front direction is 0.95, corresponding to an angle of 18.17 degrees. Thus the angle between the vector pointing to camera and the back direction is 161.83 degrees. Among the four directions, the smallest angle is the front direction, with an angle of 18.17 degrees. Thus the front side of a wooden chair with a person sitting on it is facing the camera. Therefore, the final answer is A. front.
|
A. front.
|
orientation_viewpoint
|
005ba732f68b2949.jpg
|
005bde4e3b45a6ab_29af
|
Consider the real-world 3D locations and orientations of the objects. Which side of a wooden clock with a gold trim is facing the camera?
|
front
|
left
|
back
|
right
|
[
{
"bbox_3d": [
0.7,
0.6,
4.1
],
"label": "a wooden clock with a gold trim"
}
] |
[
{
"front_dir": [
-0.1,
-0.2,
-1
],
"label": "a wooden clock with a gold trim",
"left_dir": [
-1,
0.1,
0.1
]
}
] |
A
|
To solve this problem, we first estimate the 3D location of a wooden clock with a gold trim. Then we obtain the vector pointing from the object to the camera. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a wooden clock with a gold trim, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a wooden clock with a gold trim that is facing the camera corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The location of a wooden clock with a gold trim is (0.7, 0.6, 4.1). The vector from a wooden clock with a gold trim to camera is hence (-0.7, -0.6, -4.1). The left direction of a wooden clock with a gold trim is (-1.0, 0.1, 0.1). The cosine similarity between the vector pointing to camera and the left direction is 0.08, corresponding to an angle of 85.62 degrees. Thus the angle between the vector pointing to camera and the right direction is 94.38 degrees. The front direction of a wooden clock with a gold trim is (-0.1, -0.2, -1.0). The cosine similarity between the vector pointing to camera and the front direction is 1.00, corresponding to an angle of 5.54 degrees. Thus the angle between the vector pointing to camera and the back direction is 174.46 degrees. Among the four directions, the smallest angle is the front direction, with an angle of 5.54 degrees. Thus the front side of a wooden clock with a gold trim is facing the camera. Therefore, the final answer is A. front.
|
A. front.
|
orientation_viewpoint
|
005bde4e3b45a6ab.jpg
|
005cab9a92983efa_760c
|
Consider the real-world 3D locations and orientations of the objects. Which side of a tank is facing a man in a black hat holding a cell phone?
|
front
|
left
|
back
|
right
|
[
{
"bbox_3d": [
1.1,
1.4,
3.9
],
"label": "a tank"
},
{
"bbox_3d": [
0.1,
1.1,
5.5
],
"label": "a man in a black hat holding a cell phone"
}
] |
[
{
"front_dir": [
-0.8,
0,
-0.5
],
"label": "a tank",
"left_dir": [
-0.5,
0.2,
0.8
]
}
] |
B
|
To solve this problem, we first detect the 3D locations of a tank and a man in a black hat holding a cell phone. Then we compute the vector pointing from a tank to a man in a black hat holding a cell phone. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a tank, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a tank that is facing a man in a black hat holding a cell phone corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The 3D location of a tank is (1.1, 1.4, 3.9). The 3D location of a man in a black hat holding a cell phone is (0.1, 1.1, 5.5). The vector from a tank to a man in a black hat holding a cell phone is hence (-1.0, -0.3, 1.6). The left direction of a tank is (-0.5, 0.2, 0.8). The cosine similarity between the vector pointing to a man in a black hat holding a cell phone and the left direction is 0.94, corresponding to an angle of 20.57 degrees. Thus the angle between the vector pointing to a man in a black hat holding a cell phone and the right direction is 159.43 degrees. The front direction of a tank is (-0.8, -0.0, -0.5). The cosine similarity between the vector pointing to a man in a black hat holding a cell phone and the front direction is 0.02, corresponding to an angle of 88.80 degrees. Thus the angle between the vector pointing to a man in a black hat holding a cell phone and the back direction is 91.20 degrees. Among the four directions, the smallest angle is the left direction, with an angle of 20.57 degrees. Thus the left side of a tank is facing the a man in a black hat holding a cell phone. Therefore, the final answer is B. left.
|
B. left.
|
multi_object_viewpoint_towards_object
|
005cab9a92983efa.jpg
|
005cc1a30d8029f9_1a60
|
Consider the real-world 3D orientations of the objects. What is the relationship between the orientations of a green board with a black surface and a green armchair, parallel of perpendicular to each other?
|
parallel
|
perpendicular
| null | null |
[
{
"bbox_3d": [
0.2,
1.7,
7.5
],
"label": "a green board with a black surface"
},
{
"bbox_3d": [
-0.7,
0.4,
1.9
],
"label": "a green armchair"
}
] |
[
{
"front_dir": [
0.1,
-0.1,
-1
],
"label": "a green board with a black surface",
"left_dir": [
-1,
0.1,
-0.1
]
},
{
"front_dir": [
0.4,
0,
-0.9
],
"label": "a green armchair",
"left_dir": [
-0.9,
0.1,
-0.4
]
}
] |
A
|
To solve this problem, we first detect the front directions of a green board with a black surface and a green armchair. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is closer to 0 or 180, than to 90 degrees, then the two objects are parallel to each other. Otherwise, the two objects are perpendicular to each other. The front direction of a green board with a black surface is (0.1, -0.1, -1.0). The front direction of a green armchair is (0.4, -0.0, -0.9). The cosine similarity between the two front directions is 0.94, corresponding to an angle of 20.28. The angle is closer to 0 or 180, than to 90 degrees, meaning that the two objects are parallel to each other. Therefore, the final answer is A. parallel.
|
A. parallel.
|
multi_object_parallel
|
005cc1a30d8029f9.jpg
|
005cdc41f47be514_3c16
|
Consider the real-world 3D locations and orientations of the objects. Which side of a white car with a red light on is facing a yellow sign with black writing?
|
front
|
left
|
back
|
right
|
[
{
"bbox_3d": [
-2.1,
2.1,
5.1
],
"label": "a white car with a red light on"
},
{
"bbox_3d": [
-10.4,
12.3,
32.8
],
"label": "a yellow sign with black writing"
}
] |
[
{
"front_dir": [
-0.9,
-0.1,
-0.5
],
"label": "a white car with a red light on",
"left_dir": [
-0.5,
0.1,
0.9
]
},
{
"front_dir": [
0.4,
-0.3,
-0.9
],
"label": "a yellow sign with black writing",
"left_dir": [
-0.9,
0,
-0.4
]
}
] |
B
|
To solve this problem, we first detect the 3D locations of a white car with a red light on and a yellow sign with black writing. Then we compute the vector pointing from a white car with a red light on to a yellow sign with black writing. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a white car with a red light on, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a white car with a red light on that is facing a yellow sign with black writing corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The 3D location of a white car with a red light on is (-2.1, 2.1, 5.1). The 3D location of a yellow sign with black writing is (-10.4, 12.3, 32.8). The vector from a white car with a red light on to a yellow sign with black writing is hence (-8.3, 10.2, 27.7). The left direction of a white car with a red light on is (-0.5, 0.1, 0.9). The cosine similarity between the vector pointing to a yellow sign with black writing and the left direction is 0.95, corresponding to an angle of 18.74 degrees. Thus the angle between the vector pointing to a yellow sign with black writing and the right direction is 161.26 degrees. The front direction of a white car with a red light on is (-0.9, -0.1, -0.5). The cosine similarity between the vector pointing to a yellow sign with black writing and the front direction is -0.19, corresponding to an angle of 101.16 degrees. Thus the angle between the vector pointing to a yellow sign with black writing and the back direction is 78.84 degrees. Among the four directions, the smallest angle is the left direction, with an angle of 18.74 degrees. Thus the left side of a white car with a red light on is facing the a yellow sign with black writing. Therefore, the final answer is B. left.
|
B. left.
|
multi_object_viewpoint_towards_object
|
005cdc41f47be514.jpg
|
005cdc41f47be514_af4f
|
Consider the real-world 3D orientations of the objects. Are a black car and a yellow sign with black writing facing same or similar directions, or very different directions?
|
same or similar directions
|
very different directions
| null | null |
[
{
"bbox_3d": [
-0.4,
2.3,
3.7
],
"label": "a black car"
},
{
"bbox_3d": [
-10.4,
12.3,
32.8
],
"label": "a yellow sign with black writing"
}
] |
[
{
"front_dir": [
-1,
0.1,
-0.1
],
"label": "a black car",
"left_dir": [
-0.1,
0.1,
1
]
},
{
"front_dir": [
0.4,
-0.3,
-0.9
],
"label": "a yellow sign with black writing",
"left_dir": [
-0.9,
0,
-0.4
]
}
] |
B
|
To solve this problem, we first detect the front directions of a black car and a yellow sign with black writing. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is small, then the two objects are facing same or similar directions. Otherwise, the two objects are facing very different directions. The front direction of a black car is (-1.0, 0.1, -0.1). The front direction of a yellow sign with black writing is (0.4, -0.3, -0.9). The cosine similarity between the two front directions is -0.25, corresponding to an angle of 104.45. The angle is large, meaning that the two objects are facing very different directions. Therefore, the final answer is B. very different directions.
|
B. very different directions.
|
multi_object_same_direction
|
005cdc41f47be514.jpg
|
005e4b8cc9e215d8_8341
|
Consider the real-world 3D location of the objects. Which object is closer to the camera?
|
a tall white pillar
|
a stone building with columns
| null | null |
[
{
"bbox_3d": [
0,
13.7,
46.1
],
"label": "a tall white pillar"
},
{
"bbox_3d": [
10.4,
14,
63.8
],
"label": "a stone building with columns"
}
] |
[] |
A
|
To solve this problem, we first estimate the 3D locations of a tall white pillar and a stone building with columns. Then we estimate the L2 distances from the camera to the two objects. The object with a smaller L2 distance is the one that is closer to the camera. The 3D location of a tall white pillar is (-0.0, 13.7, 46.1). The 3D location of a stone building with columns is (10.4, 14.0, 63.8). The L2 distance from the camera to a tall white pillar is 48.11. The L2 distance from the camera to a stone building with columns is 66.19. The distance to a tall white pillar is smaller. Therefore, the answer is A. a tall white pillar.
|
A. a tall white pillar.
|
location_closer_to_camera
|
005e4b8cc9e215d8.jpg
|
005e6e7428da1db1_4c3f
|
Consider the real-world 3D locations of the objects. Is a wooden carving with intricate designs directly underneath a door with a metal handle?
|
yes
|
no
| null | null |
[
{
"bbox_3d": [
0.1,
0.7,
2.3
],
"label": "a door with a metal handle"
},
{
"bbox_3d": [
-0.4,
0.4,
2.2
],
"label": "a wooden carving with intricate designs"
}
] |
[] |
B
|
To solve this problem, we first determine the 3D locations of a door with a metal handle and a wooden carving with intricate designs. Then we compute the vector pointing from a wooden carving with intricate designs to a door with a metal handle, as well as the up direction of a wooden carving with intricate designs. We estimate the cosine similarity between the vector and the up direction, which leads to the angle between the two directions. If the angle is small, then a door with a metal handle is directly above a wooden carving with intricate designs. Otherwise, then a door with a metal handle is not directly above a wooden carving with intricate designs. To solve the question, we first determine if a door with a metal handle is directly above a wooden carving with intricate designs. The 3D location of a door with a metal handle is (0.1, 0.7, 2.3). The 3D location of a wooden carving with intricate designs is (-0.4, 0.4, 2.2). The vector from a wooden carving with intricate designs to a door with a metal handle is hence (0.5, 0.3, 0.1). The up direction of a wooden carving with intricate designs is (0.0, 1.0, 0.0). The cosine similarity between the vector and the up direction is 0.53, corresponding to an angle of 57 degrees. The angle between the vector and the up direction is large, meaning that a door with a metal handle is not directly above a wooden carving with intricate designs. In other words, a wooden carving with intricate designs is not directly underneath a door with a metal handle. Therefore, the answer is B. no.
|
B. no.
|
location_above
|
005e6e7428da1db1.jpg
|
006084caa4d71586_71e0
|
Consider the real-world 3D locations and orientations of the objects. If I stand at a stone building with a door's position facing where it is facing, is a brown drum with a white stripe on the left or right of me?
|
on the left
|
on the right
| null | null |
[
{
"bbox_3d": [
-0.7,
0.9,
2.3
],
"label": "a brown drum with a white stripe"
},
{
"bbox_3d": [
2,
2.1,
6.8
],
"label": "a stone building with a door"
}
] |
[
{
"front_dir": [
0.3,
-0.2,
-0.9
],
"label": "a stone building with a door",
"left_dir": [
-1,
0,
-0.3
]
}
] |
A
|
To solve this problem, we first determine the 3D locations of a brown drum with a white stripe and a stone building with a door. Then we estimate the vector pointing from a stone building with a door to a brown drum with a white stripe, as well as the left direction of a stone building with a door. Next we compute the cosine similarities between the vector and the left direction, which leads to the angle between the vector and the fron direction. If the angle is smaller than 90 degrees, then a brown drum with a white stripe is on the left of a stone building with a door. Otherwise, a brown drum with a white stripe is behind a stone building with a door. The 3D location of a brown drum with a white stripe is (-0.7, 0.9, 2.3). The 3D location of a stone building with a door is (2.0, 2.1, 6.8). The vector from a stone building with a door to a brown drum with a white stripe is hence (-2.8, -1.2, -4.6). The left direction of a stone building with a door is (-1.0, -0.0, -0.3). The cosine similarity between the vector and the left direction is 0.71, corresponding to an angle of 44.75 degrees. The angle is smaller than 90 degrees, meaning that a brown drum with a white stripe is on the left of a stone building with a door. Therefore, the final answer is A. on the left.
|
A. on the left.
|
orientation_on_the_left
|
006084caa4d71586.jpg
|
0060a27761b0fd63_d4b3
|
Consider the real-world 3D locations of the objects. Is a white stool directly underneath a white booth?
|
yes
|
no
| null | null |
[
{
"bbox_3d": [
-0.7,
1.9,
5.8
],
"label": "a white booth"
},
{
"bbox_3d": [
0.6,
0.5,
3.3
],
"label": "a white stool"
}
] |
[] |
B
|
To solve this problem, we first determine the 3D locations of a white booth and a white stool. Then we compute the vector pointing from a white stool to a white booth, as well as the up direction of a white stool. We estimate the cosine similarity between the vector and the up direction, which leads to the angle between the two directions. If the angle is small, then a white booth is directly above a white stool. Otherwise, then a white booth is not directly above a white stool. To solve the question, we first determine if a white booth is directly above a white stool. The 3D location of a white booth is (-0.7, 1.9, 5.8). The 3D location of a white stool is (0.6, 0.5, 3.3). The vector from a white stool to a white booth is hence (-1.4, 1.3, 2.5). The up direction of a white stool is (0.1, 0.9, 0.4). The cosine similarity between the vector and the up direction is 0.66, corresponding to an angle of 48 degrees. The angle between the vector and the up direction is large, meaning that a white booth is not directly above a white stool. In other words, a white stool is not directly underneath a white booth. Therefore, the answer is B. no.
|
B. no.
|
location_above
|
0060a27761b0fd63.jpg
|
0060b370b08adbed_c376
|
Consider the real-world 3D locations of the objects. Are the a wooden bridge with snow on it and the a creek with snow on the side next to each other or far away from each other?
|
next to each other
|
far away from each other
| null | null |
[
{
"bbox_3d": [
1.6,
3.9,
23
],
"label": "a wooden bridge with snow on it"
},
{
"bbox_3d": [
-4.6,
0.8,
13.2
],
"label": "a creek with snow on the side"
}
] |
[] |
A
|
To solve this problem, we first estimate the 3D locations of a wooden bridge with snow on it and a creek with snow on the side. Then we can compute the L2 distance between the two objects. Next we estimate the rough sizes of the two objects. If the distance between the two objects is smaller or roughly the same as the object sizes, then the two objects are next to each other. Otherwise, the two objects are far away from each other. The 3D location of a wooden bridge with snow on it is (1.6, 3.9, 23.0). The 3D location of a creek with snow on the side is (-4.6, 0.8, 13.2). The L2 distance between the two objects is 12.02. The size of the a wooden bridge with snow on it is roughly 28.92. The size of the a creek with snow on the side is roughly 13.26. The distance between the two objects is smaller or roughly the same as the object sizes, meaning that they are next to each other. Therefore, the answer is A. next to each other.
|
A. next to each other.
|
location_next_to
|
0060b370b08adbed.jpg
|
00620ae31d5cf6ee_70f6
|
Consider the real-world 3D locations and orientations of the objects. Which side of a brown leather armchair is facing a wooden floor?
|
front
|
left
|
back
|
right
|
[
{
"bbox_3d": [
1.7,
0.8,
3
],
"label": "a brown leather armchair"
},
{
"bbox_3d": [
-1.1,
0.1,
2.9
],
"label": "a wooden floor"
}
] |
[
{
"front_dir": [
0.1,
-0.2,
-1
],
"label": "a brown leather armchair",
"left_dir": [
-1,
-0.2,
0
]
}
] |
B
|
To solve this problem, we first detect the 3D locations of a brown leather armchair and a wooden floor. Then we compute the vector pointing from a brown leather armchair to a wooden floor. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a brown leather armchair, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a brown leather armchair that is facing a wooden floor corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The 3D location of a brown leather armchair is (1.7, 0.8, 3.0). The 3D location of a wooden floor is (-1.1, 0.1, 2.9). The vector from a brown leather armchair to a wooden floor is hence (-2.8, -0.8, -0.1). The left direction of a brown leather armchair is (-1.0, -0.2, -0.0). The cosine similarity between the vector pointing to a wooden floor and the left direction is 0.99, corresponding to an angle of 6.25 degrees. Thus the angle between the vector pointing to a wooden floor and the right direction is 173.75 degrees. The front direction of a brown leather armchair is (0.1, -0.2, -1.0). The cosine similarity between the vector pointing to a wooden floor and the front direction is 0.01, corresponding to an angle of 89.29 degrees. Thus the angle between the vector pointing to a wooden floor and the back direction is 90.71 degrees. Among the four directions, the smallest angle is the left direction, with an angle of 6.25 degrees. Thus the left side of a brown leather armchair is facing the a wooden floor. Therefore, the final answer is B. left.
|
B. left.
|
multi_object_viewpoint_towards_object
|
00620ae31d5cf6ee.jpg
|
00627a11149243a2_d407
|
Consider the real-world 3D locations and orientations of the objects. If I stand at a large blue and white boat's position facing where it is facing, is a large bridge with a red boat in front of me or behind me?
|
in front of
|
behind
| null | null |
[
{
"bbox_3d": [
6,
35.1,
209.9
],
"label": "a large bridge with a red boat"
},
{
"bbox_3d": [
-28.1,
19.6,
137
],
"label": "a large blue and white boat"
}
] |
[
{
"front_dir": [
0.3,
-0.2,
-0.9
],
"label": "a large blue and white boat",
"left_dir": [
-1,
0,
-0.3
]
}
] |
B
|
To solve this problem, we first determine the 3D locations of a large bridge with a red boat and a large blue and white boat. Then we estimate the vector pointing from a large blue and white boat to a large bridge with a red boat, as well as the front direction of a large blue and white boat. Next we compute the cosine similarities between the vector and the front direction, which leads to the angle between the vector and the fron direction. If the angle is smaller than 90 degrees, then a large bridge with a red boat is in front of a large blue and white boat. Otherwise, a large bridge with a red boat is behind a large blue and white boat. The 3D location of a large bridge with a red boat is (6.0, 35.1, 209.9). The 3D location of a large blue and white boat is (-28.1, 19.6, 137.0). The vector from a large blue and white boat to a large bridge with a red boat is hence (34.1, 15.5, 72.9). The front direction of a large blue and white boat is (0.3, -0.2, -0.9). The cosine similarity between the vector and the front direction is -0.76, corresponding to an angle of 139.79 degrees. The angle is smaller than 90 degrees, meaning that a large bridge with a red boat is behind a large blue and white boat. Therefore, the final answer is B. behind.
|
B. behind.
|
orientation_in_front_of
|
00627a11149243a2.jpg
|
00627b61111a8412_3325
|
Consider the real-world 3D locations and orientations of the objects. Which side of a wooden bookshelf with a lot of items on it is facing a pig figurine?
|
front
|
left
|
back
|
right
|
[
{
"bbox_3d": [
0.7,
1.2,
2.1
],
"label": "a wooden bookshelf with a lot of items on it"
},
{
"bbox_3d": [
0.7,
1.2,
2.2
],
"label": "a pig figurine"
}
] |
[
{
"front_dir": [
-0.2,
-0.2,
-1
],
"label": "a wooden bookshelf with a lot of items on it",
"left_dir": [
-1,
0,
0.2
]
}
] |
C
|
To solve this problem, we first detect the 3D locations of a wooden bookshelf with a lot of items on it and a pig figurine. Then we compute the vector pointing from a wooden bookshelf with a lot of items on it to a pig figurine. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a wooden bookshelf with a lot of items on it, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a wooden bookshelf with a lot of items on it that is facing a pig figurine corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The 3D location of a wooden bookshelf with a lot of items on it is (0.7, 1.2, 2.1). The 3D location of a pig figurine is (0.7, 1.2, 2.2). The vector from a wooden bookshelf with a lot of items on it to a pig figurine is hence (0.0, 0.0, 0.1). The left direction of a wooden bookshelf with a lot of items on it is (-1.0, 0.0, 0.2). The cosine similarity between the vector pointing to a pig figurine and the left direction is 0.04, corresponding to an angle of 87.88 degrees. Thus the angle between the vector pointing to a pig figurine and the right direction is 92.12 degrees. The front direction of a wooden bookshelf with a lot of items on it is (-0.2, -0.2, -1.0). The cosine similarity between the vector pointing to a pig figurine and the front direction is -1.00, corresponding to an angle of 174.51 degrees. Thus the angle between the vector pointing to a pig figurine and the back direction is 5.49 degrees. Among the four directions, the smallest angle is the back direction, with an angle of 5.49 degrees. Thus the back side of a wooden bookshelf with a lot of items on it is facing the a pig figurine. Therefore, the final answer is C. back.
|
C. back.
|
multi_object_viewpoint_towards_object
|
00627b61111a8412.jpg
|
00630dcce737d626_51e0
|
Consider the real-world 3D locations and orientations of the objects. If I stand at a car with a lizard on it's position facing where it is facing, is a lizard with a long tail in front of me or behind me?
|
in front of
|
behind
| null | null |
[
{
"bbox_3d": [
0,
0.3,
0.7
],
"label": "a lizard with a long tail"
},
{
"bbox_3d": [
-0.1,
0.4,
0.7
],
"label": "a car with a lizard on it"
}
] |
[
{
"front_dir": [
-1,
0,
-0.2
],
"label": "a car with a lizard on it",
"left_dir": [
-0.2,
0.6,
0.8
]
}
] |
B
|
To solve this problem, we first determine the 3D locations of a lizard with a long tail and a car with a lizard on it. Then we estimate the vector pointing from a car with a lizard on it to a lizard with a long tail, as well as the front direction of a car with a lizard on it. Next we compute the cosine similarities between the vector and the front direction, which leads to the angle between the vector and the fron direction. If the angle is smaller than 90 degrees, then a lizard with a long tail is in front of a car with a lizard on it. Otherwise, a lizard with a long tail is behind a car with a lizard on it. The 3D location of a lizard with a long tail is (0.0, 0.3, 0.7). The 3D location of a car with a lizard on it is (-0.1, 0.4, 0.7). The vector from a car with a lizard on it to a lizard with a long tail is hence (0.1, -0.1, -0.0). The front direction of a car with a lizard on it is (-1.0, -0.0, -0.2). The cosine similarity between the vector and the front direction is -0.82, corresponding to an angle of 145.19 degrees. The angle is smaller than 90 degrees, meaning that a lizard with a long tail is behind a car with a lizard on it. Therefore, the final answer is B. behind.
|
B. behind.
|
orientation_in_front_of
|
00630dcce737d626.jpg
|
006385a617f76704_68aa
|
Consider the real-world 3D locations of the objects. Are the a black and white photo of a bridge and the a railroad with metal tracks next to each other or far away from each other?
|
next to each other
|
far away from each other
| null | null |
[
{
"bbox_3d": [
0,
0.7,
1.7
],
"label": "a black and white photo of a bridge"
},
{
"bbox_3d": [
0.4,
-0.2,
7.2
],
"label": "a railroad with metal tracks"
}
] |
[] |
A
|
To solve this problem, we first estimate the 3D locations of a black and white photo of a bridge and a railroad with metal tracks. Then we can compute the L2 distance between the two objects. Next we estimate the rough sizes of the two objects. If the distance between the two objects is smaller or roughly the same as the object sizes, then the two objects are next to each other. Otherwise, the two objects are far away from each other. The 3D location of a black and white photo of a bridge is (0.0, 0.7, 1.7). The 3D location of a railroad with metal tracks is (0.4, -0.2, 7.2). The L2 distance between the two objects is 5.60. The size of the a black and white photo of a bridge is roughly 0.95. The size of the a railroad with metal tracks is roughly 20.34. The distance between the two objects is smaller or roughly the same as the object sizes, meaning that they are next to each other. Therefore, the answer is A. next to each other.
|
A. next to each other.
|
location_next_to
|
006385a617f76704.jpg
|
006438bca1e353b3_0e31
|
Consider the real-world 3D locations and orientations of the objects. Which side of a silver car is facing the camera?
|
front
|
left
|
back
|
right
|
[
{
"bbox_3d": [
-6.2,
0.6,
17.4
],
"label": "a silver car"
}
] |
[
{
"front_dir": [
0.4,
-0.1,
-0.9
],
"label": "a silver car",
"left_dir": [
-0.9,
0,
-0.4
]
}
] |
A
|
To solve this problem, we first estimate the 3D location of a silver car. Then we obtain the vector pointing from the object to the camera. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a silver car, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a silver car that is facing the camera corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The location of a silver car is (-6.2, 0.6, 17.4). The vector from a silver car to camera is hence (6.2, -0.6, -17.4). The left direction of a silver car is (-0.9, -0.0, -0.4). The cosine similarity between the vector pointing to camera and the left direction is 0.08, corresponding to an angle of 85.53 degrees. Thus the angle between the vector pointing to camera and the right direction is 94.47 degrees. The front direction of a silver car is (0.4, -0.1, -0.9). The cosine similarity between the vector pointing to camera and the front direction is 1.00, corresponding to an angle of 5.15 degrees. Thus the angle between the vector pointing to camera and the back direction is 174.85 degrees. Among the four directions, the smallest angle is the front direction, with an angle of 5.15 degrees. Thus the front side of a silver car is facing the camera. Therefore, the final answer is A. front.
|
A. front.
|
orientation_viewpoint
|
006438bca1e353b3.jpg
|
006438bca1e353b3_0448
|
Consider the real-world 3D locations and orientations of the objects. Which side of a car parked on the street is facing the camera?
|
front
|
left
|
back
|
right
|
[
{
"bbox_3d": [
-5.7,
0.6,
13.5
],
"label": "a car parked on the street"
}
] |
[
{
"front_dir": [
0.5,
-0.1,
-0.9
],
"label": "a car parked on the street",
"left_dir": [
-0.9,
0,
-0.5
]
}
] |
A
|
To solve this problem, we first estimate the 3D location of a car parked on the street. Then we obtain the vector pointing from the object to the camera. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a car parked on the street, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a car parked on the street that is facing the camera corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The location of a car parked on the street is (-5.7, 0.6, 13.5). The vector from a car parked on the street to camera is hence (5.7, -0.6, -13.5). The left direction of a car parked on the street is (-0.9, -0.0, -0.5). The cosine similarity between the vector pointing to camera and the left direction is 0.08, corresponding to an angle of 85.45 degrees. Thus the angle between the vector pointing to camera and the right direction is 94.55 degrees. The front direction of a car parked on the street is (0.5, -0.1, -0.9). The cosine similarity between the vector pointing to camera and the front direction is 1.00, corresponding to an angle of 5.04 degrees. Thus the angle between the vector pointing to camera and the back direction is 174.96 degrees. Among the four directions, the smallest angle is the front direction, with an angle of 5.04 degrees. Thus the front side of a car parked on the street is facing the camera. Therefore, the final answer is A. front.
|
A. front.
|
orientation_viewpoint
|
006438bca1e353b3.jpg
|
00645acf166178e5_f0a9
|
Consider the real-world 3D locations of the objects. Which is closer to a group of men in suits, a red tie with a blue lanyard or a man wearing a suit?
|
a red tie with a blue lanyard
|
a man wearing a suit
| null | null |
[
{
"bbox_3d": [
-0.2,
2.7,
6.7
],
"label": "a group of men in suits"
},
{
"bbox_3d": [
-0.5,
1.9,
3.6
],
"label": "a red tie with a blue lanyard"
},
{
"bbox_3d": [
0.9,
1.5,
1.6
],
"label": "a man wearing a suit"
}
] |
[] |
A
|
To solve this problem, we first detect the 3D location of a group of men in suits, a red tie with a blue lanyard, and a man wearing a suit. Then we compute the L2 distances between a group of men in suits and a red tie with a blue lanyard, and between a group of men in suits and a man wearing a suit. The object that is closer to a group of men in suits is the one with a smaller distance. The 3D location of a group of men in suits is (-0.2, 2.7, 6.7). The 3D location of a red tie with a blue lanyard is (-0.5, 1.9, 3.6). The 3D location of a man wearing a suit is (0.9, 1.5, 1.6). The L2 distance between a group of men in suits and a red tie with a blue lanyard is 3.226392137534329. The L2 distance between a group of men in suits and a man wearing a suit is 5.34256792331456. Between the two distances, the distance between a group of men in suits and a red tie with a blue lanyard is smaller. Therefore, the final answer is A. a red tie with a blue lanyard.
|
A. a red tie with a blue lanyard.
|
multi_object_closer_to
|
00645acf166178e5.jpg
|
0064e0babe419863_e4d7
|
Consider the real-world 3D locations and orientations of the objects. If I stand at a person wearing a silver helmet's position facing where it is facing, is a man in a yellow shirt in front of me or behind me?
|
in front of
|
behind
| null | null |
[
{
"bbox_3d": [
-1.2,
1.8,
10.2
],
"label": "a man in a yellow shirt"
},
{
"bbox_3d": [
-0.5,
1.9,
6.1
],
"label": "a person wearing a silver helmet"
}
] |
[
{
"front_dir": [
-0.5,
-0.1,
-0.8
],
"label": "a person wearing a silver helmet",
"left_dir": [
-0.8,
0.4,
0.5
]
}
] |
B
|
To solve this problem, we first determine the 3D locations of a man in a yellow shirt and a person wearing a silver helmet. Then we estimate the vector pointing from a person wearing a silver helmet to a man in a yellow shirt, as well as the front direction of a person wearing a silver helmet. Next we compute the cosine similarities between the vector and the front direction, which leads to the angle between the vector and the fron direction. If the angle is smaller than 90 degrees, then a man in a yellow shirt is in front of a person wearing a silver helmet. Otherwise, a man in a yellow shirt is behind a person wearing a silver helmet. The 3D location of a man in a yellow shirt is (-1.2, 1.8, 10.2). The 3D location of a person wearing a silver helmet is (-0.5, 1.9, 6.1). The vector from a person wearing a silver helmet to a man in a yellow shirt is hence (-0.7, -0.1, 4.1). The front direction of a person wearing a silver helmet is (-0.5, -0.1, -0.8). The cosine similarity between the vector and the front direction is -0.74, corresponding to an angle of 137.42 degrees. The angle is smaller than 90 degrees, meaning that a man in a yellow shirt is behind a person wearing a silver helmet. Therefore, the final answer is B. behind.
|
B. behind.
|
orientation_in_front_of
|
0064e0babe419863.jpg
|
0065ac89cef9c123_13fa
|
Consider the real-world 3D orientations of the objects. What is the relationship between the orientations of a white stool with a plant on it and a white stool with a square top, parallel of perpendicular to each other?
|
parallel
|
perpendicular
| null | null |
[
{
"bbox_3d": [
0.7,
0.3,
8.1
],
"label": "a white stool with a plant on it"
},
{
"bbox_3d": [
-0.9,
0.5,
9
],
"label": "a white stool with a square top"
}
] |
[
{
"front_dir": [
0,
0.1,
-1
],
"label": "a white stool with a plant on it",
"left_dir": [
-1,
0,
0
]
},
{
"front_dir": [
0.1,
-1,
-0.2
],
"label": "a white stool with a square top",
"left_dir": [
-1,
0,
-0.1
]
}
] |
B
|
To solve this problem, we first detect the front directions of a white stool with a plant on it and a white stool with a square top. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is closer to 0 or 180, than to 90 degrees, then the two objects are parallel to each other. Otherwise, the two objects are perpendicular to each other. The front direction of a white stool with a plant on it is (-0.0, 0.1, -1.0). The front direction of a white stool with a square top is (0.1, -1.0, -0.2). The cosine similarity between the two front directions is 0.12, corresponding to an angle of 82.96. The angle is closer to 90, than to 0 or 180 degrees, meaning that the two objects are perpendicular to each other. Therefore, the final answer is B. perpendicular.
|
B. perpendicular.
|
multi_object_parallel
|
0065ac89cef9c123.jpg
|
0066f83042d2e91d_ead5
|
Consider the real-world 3D orientations of the objects. Are a white sign with black writing and a black and white street sign facing same or similar directions, or very different directions?
|
same or similar directions
|
very different directions
| null | null |
[
{
"bbox_3d": [
1.4,
2.7,
18.5
],
"label": "a white sign with black writing"
},
{
"bbox_3d": [
1.7,
4.9,
16.7
],
"label": "a black and white street sign"
}
] |
[
{
"front_dir": [
-0.3,
-0.1,
-0.9
],
"label": "a white sign with black writing",
"left_dir": [
-0.9,
0.1,
0.3
]
},
{
"front_dir": [
0,
-0.3,
-1
],
"label": "a black and white street sign",
"left_dir": [
-1,
0.1,
0
]
}
] |
A
|
To solve this problem, we first detect the front directions of a white sign with black writing and a black and white street sign. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is small, then the two objects are facing same or similar directions. Otherwise, the two objects are facing very different directions. The front direction of a white sign with black writing is (-0.3, -0.1, -0.9). The front direction of a black and white street sign is (-0.0, -0.3, -1.0). The cosine similarity between the two front directions is 0.95, corresponding to an angle of 18.76. The angle is small, meaning that the two objects are facing same or similar directions. Therefore, the final answer is A. same or similar directions.
|
A. same or similar directions.
|
multi_object_same_direction
|
0066f83042d2e91d.jpg
|
006705f39cabe8a3_3a63
|
Consider the real-world 3D locations and orientations of the objects. Which side of a white sign with a bus stop is facing the camera?
|
front
|
left
|
back
|
right
|
[
{
"bbox_3d": [
0.1,
0.9,
1.4
],
"label": "a white sign with a bus stop"
}
] |
[
{
"front_dir": [
-0.3,
-0.6,
-0.8
],
"label": "a white sign with a bus stop",
"left_dir": [
-1,
0.2,
0.2
]
}
] |
A
|
To solve this problem, we first estimate the 3D location of a white sign with a bus stop. Then we obtain the vector pointing from the object to the camera. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a white sign with a bus stop, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a white sign with a bus stop that is facing the camera corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The location of a white sign with a bus stop is (0.1, 0.9, 1.4). The vector from a white sign with a bus stop to camera is hence (-0.1, -0.9, -1.4). The left direction of a white sign with a bus stop is (-1.0, 0.2, 0.2). The cosine similarity between the vector pointing to camera and the left direction is -0.22, corresponding to an angle of 102.43 degrees. Thus the angle between the vector pointing to camera and the right direction is 77.57 degrees. The front direction of a white sign with a bus stop is (-0.3, -0.6, -0.8). The cosine similarity between the vector pointing to camera and the front direction is 0.98, corresponding to an angle of 12.58 degrees. Thus the angle between the vector pointing to camera and the back direction is 167.42 degrees. Among the four directions, the smallest angle is the front direction, with an angle of 12.58 degrees. Thus the front side of a white sign with a bus stop is facing the camera. Therefore, the final answer is A. front.
|
A. front.
|
orientation_viewpoint
|
006705f39cabe8a3.jpg
|
0067fc4ccbbf7167_182b
|
Consider the real-world 3D locations of the objects. Which is closer to a glass of alcohol, a man in a red hat or a woman in a black and white shirt?
|
a man in a red hat
|
a woman in a black and white shirt
| null | null |
[
{
"bbox_3d": [
-0.4,
1,
1.1
],
"label": "a glass of alcohol"
},
{
"bbox_3d": [
-0.6,
0.9,
2.1
],
"label": "a man in a red hat"
},
{
"bbox_3d": [
0.1,
1.3,
1.2
],
"label": "a woman in a black and white shirt"
}
] |
[] |
B
|
To solve this problem, we first detect the 3D location of a glass of alcohol, a man in a red hat, and a woman in a black and white shirt. Then we compute the L2 distances between a glass of alcohol and a man in a red hat, and between a glass of alcohol and a woman in a black and white shirt. The object that is closer to a glass of alcohol is the one with a smaller distance. The 3D location of a glass of alcohol is (-0.4, 1.0, 1.1). The 3D location of a man in a red hat is (-0.6, 0.9, 2.1). The 3D location of a woman in a black and white shirt is (0.1, 1.3, 1.2). The L2 distance between a glass of alcohol and a man in a red hat is 1.0681232397985132. The L2 distance between a glass of alcohol and a woman in a black and white shirt is 0.5978953543740753. Between the two distances, the distance between a glass of alcohol and a woman in a black and white shirt is smaller. Therefore, the final answer is B. a woman in a black and white shirt.
|
B. a woman in a black and white shirt.
|
multi_object_closer_to
|
0067fc4ccbbf7167.jpg
|
006873f336a33f4e_dfba
|
Consider the real-world 3D locations of the objects. Is a long branch with a twig on it directly underneath a bird with blue and green feathers?
|
yes
|
no
| null | null |
[
{
"bbox_3d": [
0.2,
0.7,
4.4
],
"label": "a bird with blue and green feathers"
},
{
"bbox_3d": [
0.1,
0.2,
3.6
],
"label": "a long branch with a twig on it"
}
] |
[] |
B
|
To solve this problem, we first determine the 3D locations of a bird with blue and green feathers and a long branch with a twig on it. Then we compute the vector pointing from a long branch with a twig on it to a bird with blue and green feathers, as well as the up direction of a long branch with a twig on it. We estimate the cosine similarity between the vector and the up direction, which leads to the angle between the two directions. If the angle is small, then a bird with blue and green feathers is directly above a long branch with a twig on it. Otherwise, then a bird with blue and green feathers is not directly above a long branch with a twig on it. To solve the question, we first determine if a bird with blue and green feathers is directly above a long branch with a twig on it. The 3D location of a bird with blue and green feathers is (0.2, 0.7, 4.4). The 3D location of a long branch with a twig on it is (0.1, 0.2, 3.6). The vector from a long branch with a twig on it to a bird with blue and green feathers is hence (0.1, 0.5, 0.7). The up direction of a long branch with a twig on it is (0.0, 1.0, 0.0). The cosine similarity between the vector and the up direction is 0.59, corresponding to an angle of 53 degrees. The angle between the vector and the up direction is large, meaning that a bird with blue and green feathers is not directly above a long branch with a twig on it. In other words, a long branch with a twig on it is not directly underneath a bird with blue and green feathers. Therefore, the answer is B. no.
|
B. no.
|
location_above
|
006873f336a33f4e.jpg
|
006abd52cd339c78_84a7
|
Consider the real-world 3D locations and orientations of the objects. Which side of a black wooden bench is facing the camera?
|
front
|
left
|
back
|
right
|
[
{
"bbox_3d": [
2.1,
0.6,
3.5
],
"label": "a black wooden bench"
}
] |
[
{
"front_dir": [
-0.4,
0,
-0.9
],
"label": "a black wooden bench",
"left_dir": [
-0.9,
-0.1,
0.4
]
}
] |
A
|
To solve this problem, we first estimate the 3D location of a black wooden bench. Then we obtain the vector pointing from the object to the camera. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a black wooden bench, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a black wooden bench that is facing the camera corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The location of a black wooden bench is (2.1, 0.6, 3.5). The vector from a black wooden bench to camera is hence (-2.1, -0.6, -3.5). The left direction of a black wooden bench is (-0.9, -0.1, 0.4). The cosine similarity between the vector pointing to camera and the left direction is 0.11, corresponding to an angle of 83.89 degrees. Thus the angle between the vector pointing to camera and the right direction is 96.11 degrees. The front direction of a black wooden bench is (-0.4, 0.0, -0.9). The cosine similarity between the vector pointing to camera and the front direction is 0.98, corresponding to an angle of 10.36 degrees. Thus the angle between the vector pointing to camera and the back direction is 169.64 degrees. Among the four directions, the smallest angle is the front direction, with an angle of 10.36 degrees. Thus the front side of a black wooden bench is facing the camera. Therefore, the final answer is A. front.
|
A. front.
|
orientation_viewpoint
|
006abd52cd339c78.jpg
|
006b01565ffa7ef6_55ea
|
Consider the real-world 3D locations of the objects. Which is closer to a white pole with a sign on it, a metal rail or a black and white road?
|
a metal rail
|
a black and white road
| null | null |
[
{
"bbox_3d": [
0.6,
3.6,
10.5
],
"label": "a white pole with a sign on it"
},
{
"bbox_3d": [
3.7,
0.7,
14.8
],
"label": "a metal rail"
},
{
"bbox_3d": [
-0.2,
0.1,
3.2
],
"label": "a black and white road"
}
] |
[] |
A
|
To solve this problem, we first detect the 3D location of a white pole with a sign on it, a metal rail, and a black and white road. Then we compute the L2 distances between a white pole with a sign on it and a metal rail, and between a white pole with a sign on it and a black and white road. The object that is closer to a white pole with a sign on it is the one with a smaller distance. The 3D location of a white pole with a sign on it is (0.6, 3.6, 10.5). The 3D location of a metal rail is (3.7, 0.7, 14.8). The 3D location of a black and white road is (-0.2, 0.1, 3.2). The L2 distance between a white pole with a sign on it and a metal rail is 6.043988299913951. The L2 distance between a white pole with a sign on it and a black and white road is 8.131345502625894. Between the two distances, the distance between a white pole with a sign on it and a metal rail is smaller. Therefore, the final answer is A. a metal rail.
|
A. a metal rail.
|
multi_object_closer_to
|
006b01565ffa7ef6.jpg
|
006c2bd42fda3598_a210
|
Consider the real-world 3D orientations of the objects. Are a black car and a motorbike with a black and orange color scheme facing same or similar directions, or very different directions?
|
same or similar directions
|
very different directions
| null | null |
[
{
"bbox_3d": [
1.6,
0.9,
5.6
],
"label": "a black car"
},
{
"bbox_3d": [
-0.2,
1.6,
2.8
],
"label": "a motorbike with a black and orange color scheme"
}
] |
[
{
"front_dir": [
-0.3,
-0.1,
1
],
"label": "a black car",
"left_dir": [
1,
0,
0.3
]
},
{
"front_dir": [
0.7,
-0.1,
-0.7
],
"label": "a motorbike with a black and orange color scheme",
"left_dir": [
-0.7,
-0.1,
-0.7
]
}
] |
B
|
To solve this problem, we first detect the front directions of a black car and a motorbike with a black and orange color scheme. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is small, then the two objects are facing same or similar directions. Otherwise, the two objects are facing very different directions. The front direction of a black car is (-0.3, -0.1, 1.0). The front direction of a motorbike with a black and orange color scheme is (0.7, -0.1, -0.7). The cosine similarity between the two front directions is -0.87, corresponding to an angle of 150.65. The angle is large, meaning that the two objects are facing very different directions. Therefore, the final answer is B. very different directions.
|
B. very different directions.
|
multi_object_same_direction
|
006c2bd42fda3598.jpg
|
006d199d383d7976_075f
|
Consider the real-world 3D locations and orientations of the objects. Which side of a white van is facing a sheep with a long tail?
|
front
|
left
|
back
|
right
|
[
{
"bbox_3d": [
4.6,
2.2,
25.6
],
"label": "a white van"
},
{
"bbox_3d": [
-3.7,
0.7,
24.5
],
"label": "a sheep with a long tail"
}
] |
[
{
"front_dir": [
1,
0,
-0.1
],
"label": "a white van",
"left_dir": [
-0.1,
-0.1,
-1
]
}
] |
C
|
To solve this problem, we first detect the 3D locations of a white van and a sheep with a long tail. Then we compute the vector pointing from a white van to a sheep with a long tail. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a white van, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a white van that is facing a sheep with a long tail corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The 3D location of a white van is (4.6, 2.2, 25.6). The 3D location of a sheep with a long tail is (-3.7, 0.7, 24.5). The vector from a white van to a sheep with a long tail is hence (-8.4, -1.5, -1.2). The left direction of a white van is (-0.1, -0.1, -1.0). The cosine similarity between the vector pointing to a sheep with a long tail and the left direction is 0.24, corresponding to an angle of 75.88 degrees. Thus the angle between the vector pointing to a sheep with a long tail and the right direction is 104.12 degrees. The front direction of a white van is (1.0, 0.0, -0.1). The cosine similarity between the vector pointing to a sheep with a long tail and the front direction is -0.96, corresponding to an angle of 164.25 degrees. Thus the angle between the vector pointing to a sheep with a long tail and the back direction is 15.75 degrees. Among the four directions, the smallest angle is the back direction, with an angle of 15.75 degrees. Thus the back side of a white van is facing the a sheep with a long tail. Therefore, the final answer is C. back.
|
C. back.
|
multi_object_viewpoint_towards_object
|
006d199d383d7976.jpg
|
006d1e500c04c430_f7ae
|
Consider the real-world 3D locations of the objects. Which object has a higher location?
|
a man in blue shorts and a white shirt
|
a group of people standing
| null | null |
[
{
"bbox_3d": [
0,
1.2,
3
],
"label": "a man in blue shorts and a white shirt"
},
{
"bbox_3d": [
-14.6,
29.4,
113.1
],
"label": "a group of people standing"
}
] |
[] |
B
|
To solve this problem, we first estimate the 3D heights of the two objects. The object with a larger height value is at a higher location. The object with a smaller height value is at a lower location. The 3D height of a man in blue shorts and a white shirt is 3.1. The 3D height of a group of people standing is 82.7. The 3D height of a group of people standing is larger, meaning that the location of a group of people standing is higher. Therefore, the answer is B. a group of people standing.
|
B. a group of people standing.
|
height_higher
|
006d1e500c04c430.jpg
|
006da7d8e0ea820a_b019
|
Consider the real-world 3D orientations of the objects. What is the relationship between the orientations of a large boat in the water and a large boat is floating on the water, parallel of perpendicular to each other?
|
parallel
|
perpendicular
| null | null |
[
{
"bbox_3d": [
8.2,
-8.8,
54.2
],
"label": "a large boat in the water"
},
{
"bbox_3d": [
-20.6,
-9.7,
118.8
],
"label": "a large boat is floating on the water"
}
] |
[
{
"front_dir": [
-0.1,
0.1,
-1
],
"label": "a large boat in the water",
"left_dir": [
-1,
0.1,
0.1
]
},
{
"front_dir": [
0.3,
0.2,
-1
],
"label": "a large boat is floating on the water",
"left_dir": [
-1,
0.1,
-0.2
]
}
] |
A
|
To solve this problem, we first detect the front directions of a large boat in the water and a large boat is floating on the water. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is closer to 0 or 180, than to 90 degrees, then the two objects are parallel to each other. Otherwise, the two objects are perpendicular to each other. The front direction of a large boat in the water is (-0.1, 0.1, -1.0). The front direction of a large boat is floating on the water is (0.3, 0.2, -1.0). The cosine similarity between the two front directions is 0.94, corresponding to an angle of 19.46. The angle is closer to 0 or 180, than to 90 degrees, meaning that the two objects are parallel to each other. Therefore, the final answer is A. parallel.
|
A. parallel.
|
multi_object_parallel
|
006da7d8e0ea820a.jpg
|
006e15b7e76cc22f_1e30
|
Consider the real-world 3D locations of the objects. Which object has a lower location?
|
a white cake with brown frosting
|
a child in a white shirt
| null | null |
[
{
"bbox_3d": [
0,
0.3,
2.1
],
"label": "a white cake with brown frosting"
},
{
"bbox_3d": [
-0.7,
0.4,
2
],
"label": "a child in a white shirt"
}
] |
[] |
A
|
To solve this problem, we first estimate the 3D heights of the two objects. The object with a larger height value is at a higher location. The object with a smaller height value is at a lower location. The 3D height of a white cake with brown frosting is 0.4. The 3D height of a child in a white shirt is 0.9. The 3D height of a child in a white shirt is larger, meaning that the location of a child in a white shirt is higher. In other words, the location of a white cake with brown frosting is lower. Therefore, the answer is A. a white cake with brown frosting.
|
A. a white cake with brown frosting
|
height_higher
|
006e15b7e76cc22f.jpg
|
006e34c92c516fe1_c819
|
Consider the real-world 3D locations and orientations of the objects. If I stand at a red handcart with a plant in it's position facing where it is facing, is a man wearing a white hat in front of me or behind me?
|
in front of
|
behind
| null | null |
[
{
"bbox_3d": [
-1.3,
0.7,
4.7
],
"label": "a man wearing a white hat"
},
{
"bbox_3d": [
-0.7,
0.1,
2.3
],
"label": "a red handcart with a plant in it"
}
] |
[
{
"front_dir": [
0.2,
0.3,
-0.9
],
"label": "a red handcart with a plant in it",
"left_dir": [
-1,
0,
-0.2
]
}
] |
B
|
To solve this problem, we first determine the 3D locations of a man wearing a white hat and a red handcart with a plant in it. Then we estimate the vector pointing from a red handcart with a plant in it to a man wearing a white hat, as well as the front direction of a red handcart with a plant in it. Next we compute the cosine similarities between the vector and the front direction, which leads to the angle between the vector and the fron direction. If the angle is smaller than 90 degrees, then a man wearing a white hat is in front of a red handcart with a plant in it. Otherwise, a man wearing a white hat is behind a red handcart with a plant in it. The 3D location of a man wearing a white hat is (-1.3, 0.7, 4.7). The 3D location of a red handcart with a plant in it is (-0.7, 0.1, 2.3). The vector from a red handcart with a plant in it to a man wearing a white hat is hence (-0.7, 0.6, 2.4). The front direction of a red handcart with a plant in it is (0.2, 0.3, -0.9). The cosine similarity between the vector and the front direction is -0.85, corresponding to an angle of 148.27 degrees. The angle is smaller than 90 degrees, meaning that a man wearing a white hat is behind a red handcart with a plant in it. Therefore, the final answer is B. behind.
|
B. behind.
|
orientation_in_front_of
|
006e34c92c516fe1.jpg
|
006e63a5b3d15712_7cff
|
Consider the real-world 3D locations and orientations of the objects. Which side of a wicker chair is facing a brown chair with a black metal frame?
|
front
|
left
|
back
|
right
|
[
{
"bbox_3d": [
1.4,
0.7,
9.3
],
"label": "a wicker chair"
},
{
"bbox_3d": [
-1.5,
1,
10.7
],
"label": "a brown chair with a black metal frame"
}
] |
[
{
"front_dir": [
0.8,
0,
-0.6
],
"label": "a wicker chair",
"left_dir": [
-0.6,
0.1,
-0.8
]
},
{
"front_dir": [
1,
-0.1,
0.2
],
"label": "a brown chair with a black metal frame",
"left_dir": [
0.2,
0.1,
-1
]
}
] |
C
|
To solve this problem, we first detect the 3D locations of a wicker chair and a brown chair with a black metal frame. Then we compute the vector pointing from a wicker chair to a brown chair with a black metal frame. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a wicker chair, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a wicker chair that is facing a brown chair with a black metal frame corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The 3D location of a wicker chair is (1.4, 0.7, 9.3). The 3D location of a brown chair with a black metal frame is (-1.5, 1.0, 10.7). The vector from a wicker chair to a brown chair with a black metal frame is hence (-2.9, 0.3, 1.5). The left direction of a wicker chair is (-0.6, 0.1, -0.8). The cosine similarity between the vector pointing to a brown chair with a black metal frame and the left direction is 0.25, corresponding to an angle of 75.63 degrees. Thus the angle between the vector pointing to a brown chair with a black metal frame and the right direction is 104.37 degrees. The front direction of a wicker chair is (0.8, 0.0, -0.6). The cosine similarity between the vector pointing to a brown chair with a black metal frame and the front direction is -0.96, corresponding to an angle of 164.71 degrees. Thus the angle between the vector pointing to a brown chair with a black metal frame and the back direction is 15.29 degrees. Among the four directions, the smallest angle is the back direction, with an angle of 15.29 degrees. Thus the back side of a wicker chair is facing the a brown chair with a black metal frame. Therefore, the final answer is C. back.
|
C. back.
|
multi_object_viewpoint_towards_object
|
006e63a5b3d15712.jpg
|
006eb0d7662cd8f4_420d
|
Consider the real-world 3D locations of the objects. Is a large building with a lot of windows directly above a bench in a atrium?
|
yes
|
no
| null | null |
[
{
"bbox_3d": [
1.5,
4.9,
12.3
],
"label": "a large building with a lot of windows"
},
{
"bbox_3d": [
0.2,
0.6,
4.4
],
"label": "a bench in a atrium"
}
] |
[] |
B
|
To solve this problem, we first determine the 3D locations of a large building with a lot of windows and a bench in a atrium. Then we compute the vector pointing from a bench in a atrium to a large building with a lot of windows, as well as the up direction of a bench in a atrium. We estimate the cosine similarity between the vector and the up direction, which leads to the angle between the two directions. If the angle is small, then a large building with a lot of windows is directly above a bench in a atrium. Otherwise, then a large building with a lot of windows is not directly above a bench in a atrium. The 3D location of a large building with a lot of windows is (1.5, 4.9, 12.3). The 3D location of a bench in a atrium is (0.2, 0.6, 4.4). The vector from a bench in a atrium to a large building with a lot of windows is hence (1.3, 4.3, 7.8). The up direction of a bench in a atrium is (0.0, 1.0, 0.0). The cosine similarity between the vector and the up direction is 0.48, corresponding to an angle of 61 degrees. The angle between the vector and the up direction is large, meaning that a large building with a lot of windows is not directly above a bench in a atrium. Therefore, the answer is B. no.
|
B. no.
|
location_above
|
006eb0d7662cd8f4.jpg
|
006f761aa7a0430d_273b
|
Consider the real-world 3D orientations of the objects. What is the relationship between the orientations of a chair with a wooden frame and a black seat and a gray chair, parallel of perpendicular to each other?
|
parallel
|
perpendicular
| null | null |
[
{
"bbox_3d": [
2.8,
0.4,
13.6
],
"label": "a chair with a wooden frame and a black seat"
},
{
"bbox_3d": [
0.9,
0.5,
13.7
],
"label": "a gray chair"
}
] |
[
{
"front_dir": [
-0.1,
0.1,
-1
],
"label": "a chair with a wooden frame and a black seat",
"left_dir": [
-1,
0.1,
0.1
]
},
{
"front_dir": [
-1,
0,
0.1
],
"label": "a gray chair",
"left_dir": [
0.1,
0.1,
1
]
}
] |
B
|
To solve this problem, we first detect the front directions of a chair with a wooden frame and a black seat and a gray chair. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is closer to 0 or 180, than to 90 degrees, then the two objects are parallel to each other. Otherwise, the two objects are perpendicular to each other. The front direction of a chair with a wooden frame and a black seat is (-0.1, 0.1, -1.0). The front direction of a gray chair is (-1.0, -0.0, 0.1). The cosine similarity between the two front directions is -0.01, corresponding to an angle of 90.70. The angle is closer to 90, than to 0 or 180 degrees, meaning that the two objects are perpendicular to each other. Therefore, the final answer is B. perpendicular.
|
B. perpendicular.
|
multi_object_parallel
|
006f761aa7a0430d.jpg
|
006f9eddc9634c9c_e609
|
Consider the real-world 3D locations of the objects. Are the a dock with a yellow and black pole and the a large white boat next to each other or far away from each other?
|
next to each other
|
far away from each other
| null | null |
[
{
"bbox_3d": [
-5.9,
2.5,
15
],
"label": "a dock with a yellow and black pole"
},
{
"bbox_3d": [
-3.1,
7.1,
48.6
],
"label": "a large white boat"
}
] |
[] |
A
|
To solve this problem, we first estimate the 3D locations of a dock with a yellow and black pole and a large white boat. Then we can compute the L2 distance between the two objects. Next we estimate the rough sizes of the two objects. If the distance between the two objects is smaller or roughly the same as the object sizes, then the two objects are next to each other. Otherwise, the two objects are far away from each other. The 3D location of a dock with a yellow and black pole is (-5.9, 2.5, 15.0). The 3D location of a large white boat is (-3.1, 7.1, 48.6). The L2 distance between the two objects is 34.09. The size of the a dock with a yellow and black pole is roughly 21.69. The size of the a large white boat is roughly 184.83. The distance between the two objects is smaller or roughly the same as the object sizes, meaning that they are next to each other. Therefore, the answer is A. next to each other.
|
A. next to each other.
|
location_next_to
|
006f9eddc9634c9c.jpg
|
007003e7f35ec6e5_ce55
|
Consider the real-world 3D location of the objects. Which object is closer to the camera?
|
a skeleton with a cigarette in its mouth
|
a basement with a tile floor
| null | null |
[
{
"bbox_3d": [
-0.1,
2.5,
5
],
"label": "a skeleton with a cigarette in its mouth"
},
{
"bbox_3d": [
0.7,
0.2,
4.3
],
"label": "a basement with a tile floor"
}
] |
[] |
B
|
To solve this problem, we first estimate the 3D locations of a skeleton with a cigarette in its mouth and a basement with a tile floor. Then we estimate the L2 distances from the camera to the two objects. The object with a smaller L2 distance is the one that is closer to the camera. The 3D location of a skeleton with a cigarette in its mouth is (-0.1, 2.5, 5.0). The 3D location of a basement with a tile floor is (0.7, 0.2, 4.3). The L2 distance from the camera to a skeleton with a cigarette in its mouth is 5.57. The L2 distance from the camera to a basement with a tile floor is 4.38. The distance to a basement with a tile floor is smaller. Therefore, the answer is B. a basement with a tile floor.
|
B. a basement with a tile floor.
|
location_closer_to_camera
|
007003e7f35ec6e5.jpg
|
00702f9bbe4d86a5_8b7d
|
Consider the real-world 3D locations and orientations of the objects. Which side of a man in a plaid shirt is pointing at a black car is facing a man in a plaid shirt?
|
front
|
left
|
back
|
right
|
[
{
"bbox_3d": [
0.3,
1.1,
3.6
],
"label": "a man in a plaid shirt is pointing at a black car"
},
{
"bbox_3d": [
-0.6,
1.3,
2.4
],
"label": "a man in a plaid shirt"
}
] |
[
{
"front_dir": [
0.8,
0.1,
-0.6
],
"label": "a man in a plaid shirt is pointing at a black car",
"left_dir": [
-0.6,
0.1,
-0.8
]
}
] |
B
|
To solve this problem, we first detect the 3D locations of a man in a plaid shirt is pointing at a black car and a man in a plaid shirt. Then we compute the vector pointing from a man in a plaid shirt is pointing at a black car to a man in a plaid shirt. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a man in a plaid shirt is pointing at a black car, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a man in a plaid shirt is pointing at a black car that is facing a man in a plaid shirt corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The 3D location of a man in a plaid shirt is pointing at a black car is (0.3, 1.1, 3.6). The 3D location of a man in a plaid shirt is (-0.6, 1.3, 2.4). The vector from a man in a plaid shirt is pointing at a black car to a man in a plaid shirt is hence (-0.9, 0.2, -1.2). The left direction of a man in a plaid shirt is pointing at a black car is (-0.6, 0.1, -0.8). The cosine similarity between the vector pointing to a man in a plaid shirt and the left direction is 1.00, corresponding to an angle of 5.24 degrees. Thus the angle between the vector pointing to a man in a plaid shirt and the right direction is 174.76 degrees. The front direction of a man in a plaid shirt is pointing at a black car is (0.8, 0.1, -0.6). The cosine similarity between the vector pointing to a man in a plaid shirt and the front direction is -0.04, corresponding to an angle of 92.53 degrees. Thus the angle between the vector pointing to a man in a plaid shirt and the back direction is 87.47 degrees. Among the four directions, the smallest angle is the left direction, with an angle of 5.24 degrees. Thus the left side of a man in a plaid shirt is pointing at a black car is facing the a man in a plaid shirt. Therefore, the final answer is B. left.
|
B. left.
|
multi_object_viewpoint_towards_object
|
00702f9bbe4d86a5.jpg
|
0070de5ed81e6f6e_4065
|
Consider the real-world 3D locations and orientations of the objects. Which object is a wooden chair with a black cushion facing towards, a restaurant with a large window or the a white napkin on a plate?
|
a restaurant with a large window
|
a white napkin on a plate
| null | null |
[
{
"bbox_3d": [
1.9,
0.9,
3.3
],
"label": "a wooden chair with a black cushion"
},
{
"bbox_3d": [
0.5,
1.5,
3.1
],
"label": "a restaurant with a large window"
},
{
"bbox_3d": [
0.9,
1,
1.9
],
"label": "a white napkin on a plate"
}
] |
[
{
"front_dir": [
-0.4,
0.1,
-0.9
],
"label": "a wooden chair with a black cushion",
"left_dir": [
-0.9,
0,
0.4
]
}
] |
B
|
To solve this problem, we first detect the 3D location of a wooden chair with a black cushion, a restaurant with a large window, and a white napkin on a plate. Then we compute the cosine similarities between the front direction of a wooden chair with a black cushion and the vectors from a wooden chair with a black cushion to the other two objects. We can estimate the angles from the cosine similarities, and the object with a smaller angle is the object that a wooden chair with a black cushion is facing towards. The 3D location of a wooden chair with a black cushion is (1.9, 0.9, 3.3). The 3D location of a restaurant with a large window is (0.5, 1.5, 3.1). The 3D location of a white napkin on a plate is (0.9, 1.0, 1.9). The front direction of a wooden chair with a black cushion is (-0.4, 0.1, -0.9). First we consider if a wooden chair with a black cushion is facing towards the a restaurant with a large window. The vector from a wooden chair with a black cushion to a restaurant with a large window is (-1.4, 0.6, -0.2). The cosine similarity between the front direction and the vector is 0.52, corresponding to an angle of 58.46 degrees. First we consider if a wooden chair with a black cushion is facing towards the a white napkin on a plate. The vector from a wooden chair with a black cushion to a white napkin on a plate is (-1.0, 0.1, -1.5). The cosine similarity between the front direction and the vector is 0.99, corresponding to an angle of 8.93 degrees. We find that the angle between the front direction and a white napkin on a plate is smaller. Therefore, the final answer is B. a white napkin on a plate.
|
B. a white napkin on a plate.
|
multi_object_facing
|
0070de5ed81e6f6e.jpg
|
00713c8868c0fd83_f5d4
|
Consider the real-world 3D locations and orientations of the objects. Which side of a wooden bench with two people sitting on it is facing the camera?
|
front
|
left
|
back
|
right
|
[
{
"bbox_3d": [
-0.9,
0.9,
2.8
],
"label": "a wooden bench with two people sitting on it"
}
] |
[
{
"front_dir": [
0.4,
0.1,
-0.9
],
"label": "a wooden bench with two people sitting on it",
"left_dir": [
-0.9,
0.1,
-0.3
]
}
] |
A
|
To solve this problem, we first estimate the 3D location of a wooden bench with two people sitting on it. Then we obtain the vector pointing from the object to the camera. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a wooden bench with two people sitting on it, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a wooden bench with two people sitting on it that is facing the camera corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The location of a wooden bench with two people sitting on it is (-0.9, 0.9, 2.8). The vector from a wooden bench with two people sitting on it to camera is hence (0.9, -0.9, -2.8). The left direction of a wooden bench with two people sitting on it is (-0.9, 0.1, -0.3). The cosine similarity between the vector pointing to camera and the left direction is 0.00, corresponding to an angle of 89.91 degrees. Thus the angle between the vector pointing to camera and the right direction is 90.09 degrees. The front direction of a wooden bench with two people sitting on it is (0.4, 0.1, -0.9). The cosine similarity between the vector pointing to camera and the front direction is 0.91, corresponding to an angle of 24.97 degrees. Thus the angle between the vector pointing to camera and the back direction is 155.03 degrees. Among the four directions, the smallest angle is the front direction, with an angle of 24.97 degrees. Thus the front side of a wooden bench with two people sitting on it is facing the camera. Therefore, the final answer is A. front.
|
A. front.
|
orientation_viewpoint
|
00713c8868c0fd83.jpg
|
0071902883696658_d90a
|
Consider the real-world 3D locations and orientations of the objects. Which side of a wooden pier with a boat is facing a rainbow in the sky?
|
front
|
left
|
back
|
right
|
[
{
"bbox_3d": [
-2.8,
0.5,
18
],
"label": "a wooden pier with a boat"
},
{
"bbox_3d": [
-59.3,
20.6,
163
],
"label": "a rainbow in the sky"
}
] |
[
{
"front_dir": [
-0.1,
0.1,
1
],
"label": "a wooden pier with a boat",
"left_dir": [
1,
0.1,
0.1
]
}
] |
A
|
To solve this problem, we first detect the 3D locations of a wooden pier with a boat and a rainbow in the sky. Then we compute the vector pointing from a wooden pier with a boat to a rainbow in the sky. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a wooden pier with a boat, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a wooden pier with a boat that is facing a rainbow in the sky corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The 3D location of a wooden pier with a boat is (-2.8, 0.5, 18.0). The 3D location of a rainbow in the sky is (-59.3, 20.6, 163.0). The vector from a wooden pier with a boat to a rainbow in the sky is hence (-56.4, 20.0, 144.9). The left direction of a wooden pier with a boat is (1.0, 0.1, 0.1). The cosine similarity between the vector pointing to a rainbow in the sky and the left direction is -0.26, corresponding to an angle of 104.78 degrees. Thus the angle between the vector pointing to a rainbow in the sky and the right direction is 75.22 degrees. The front direction of a wooden pier with a boat is (-0.1, 0.1, 1.0). The cosine similarity between the vector pointing to a rainbow in the sky and the front direction is 0.97, corresponding to an angle of 15.14 degrees. Thus the angle between the vector pointing to a rainbow in the sky and the back direction is 164.86 degrees. Among the four directions, the smallest angle is the front direction, with an angle of 15.14 degrees. Thus the front side of a wooden pier with a boat is facing the a rainbow in the sky. Therefore, the final answer is A. front.
|
A. front.
|
multi_object_viewpoint_towards_object
|
0071902883696658.jpg
|
0071b1d4f9203caf_4df7
|
Consider the real-world 3D locations of the objects. Which is closer to a boat in the water, a black paddle with a white sticker or a group of people in red hats on a boat?
|
a black paddle with a white sticker
|
a group of people in red hats on a boat
| null | null |
[
{
"bbox_3d": [
5.7,
1.7,
29
],
"label": "a boat in the water"
},
{
"bbox_3d": [
-1,
0.5,
8.4
],
"label": "a black paddle with a white sticker"
},
{
"bbox_3d": [
0.9,
0.4,
14.2
],
"label": "a group of people in red hats on a boat"
}
] |
[] |
B
|
To solve this problem, we first detect the 3D location of a boat in the water, a black paddle with a white sticker, and a group of people in red hats on a boat. Then we compute the L2 distances between a boat in the water and a black paddle with a white sticker, and between a boat in the water and a group of people in red hats on a boat. The object that is closer to a boat in the water is the one with a smaller distance. The 3D location of a boat in the water is (5.7, 1.7, 29.0). The 3D location of a black paddle with a white sticker is (-1.0, 0.5, 8.4). The 3D location of a group of people in red hats on a boat is (0.9, 0.4, 14.2). The L2 distance between a boat in the water and a black paddle with a white sticker is 21.680647842972615. The L2 distance between a boat in the water and a group of people in red hats on a boat is 15.601366162035953. Between the two distances, the distance between a boat in the water and a group of people in red hats on a boat is smaller. Therefore, the final answer is B. a group of people in red hats on a boat.
|
B. a group of people in red hats on a boat.
|
multi_object_closer_to
|
0071b1d4f9203caf.jpg
|
0072bf68a76b4842_b37f
|
Consider the real-world 3D locations of the objects. Are the a city street with a blue car and the a silver car with a license plate next to each other or far away from each other?
|
next to each other
|
far away from each other
| null | null |
[
{
"bbox_3d": [
-10.6,
8,
47.9
],
"label": "a city street with a blue car"
},
{
"bbox_3d": [
-11.1,
2.4,
36.8
],
"label": "a silver car with a license plate"
}
] |
[] |
A
|
To solve this problem, we first estimate the 3D locations of a city street with a blue car and a silver car with a license plate. Then we can compute the L2 distance between the two objects. Next we estimate the rough sizes of the two objects. If the distance between the two objects is smaller or roughly the same as the object sizes, then the two objects are next to each other. Otherwise, the two objects are far away from each other. The 3D location of a city street with a blue car is (-10.6, 8.0, 47.9). The 3D location of a silver car with a license plate is (-11.1, 2.4, 36.8). The L2 distance between the two objects is 12.47. The size of the a city street with a blue car is roughly 14.19. The size of the a silver car with a license plate is roughly 34.37. The distance between the two objects is smaller or roughly the same as the object sizes, meaning that they are next to each other. Therefore, the answer is A. next to each other.
|
A. next to each other.
|
location_next_to
|
0072bf68a76b4842.jpg
|
0072bf68a76b4842_779d
|
Consider the real-world 3D orientations of the objects. Are a car with a chrome bumper and a car with a blue stripe facing same or similar directions, or very different directions?
|
same or similar directions
|
very different directions
| null | null |
[
{
"bbox_3d": [
-19.2,
3.7,
55.3
],
"label": "a car with a chrome bumper"
},
{
"bbox_3d": [
-24.7,
3.7,
63.2
],
"label": "a car with a blue stripe"
}
] |
[
{
"front_dir": [
0.7,
-0.1,
-0.7
],
"label": "a car with a chrome bumper",
"left_dir": [
-0.7,
-0.1,
-0.7
]
},
{
"front_dir": [
0.4,
-0.1,
-0.9
],
"label": "a car with a blue stripe",
"left_dir": [
-0.9,
-0.1,
-0.4
]
}
] |
A
|
To solve this problem, we first detect the front directions of a car with a chrome bumper and a car with a blue stripe. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is small, then the two objects are facing same or similar directions. Otherwise, the two objects are facing very different directions. The front direction of a car with a chrome bumper is (0.7, -0.1, -0.7). The front direction of a car with a blue stripe is (0.4, -0.1, -0.9). The cosine similarity between the two front directions is 0.96, corresponding to an angle of 15.72. The angle is small, meaning that the two objects are facing same or similar directions. Therefore, the final answer is A. same or similar directions.
|
A. same or similar directions.
|
multi_object_same_direction
|
0072bf68a76b4842.jpg
|
007300b672945e7c_9ffa
|
Consider the real-world 3D location of the objects. Which object is further away from the camera?
|
a woman in a black shirt sitting at a table
|
a microphone on a table
| null | null |
[
{
"bbox_3d": [
0.1,
0.3,
1.8
],
"label": "a woman in a black shirt sitting at a table"
},
{
"bbox_3d": [
0.2,
0.3,
1.3
],
"label": "a microphone on a table"
}
] |
[] |
A
|
To solve this problem, we first estimate the 3D locations of a woman in a black shirt sitting at a table and a microphone on a table. Then we estimate the L2 distances from the camera to the two objects. The object with a smaller L2 distance is the one that is closer to the camera. The 3D location of a woman in a black shirt sitting at a table is (0.1, 0.3, 1.8). The 3D location of a microphone on a table is (0.2, 0.3, 1.3). The L2 distance from the camera to a woman in a black shirt sitting at a table is 1.80. The L2 distance from the camera to a microphone on a table is 1.40. The distance to a woman in a black shirt sitting at a table is larger. Therefore, the answer is A. a woman in a black shirt sitting at a table.
|
A. a woman in a black shirt sitting at a table.
|
location_closer_to_camera
|
007300b672945e7c.jpg
|
007390fcb5346eff_93ba
|
Consider the real-world 3D locations and orientations of the objects. Which object is a black boat facing towards, a large body of water or the a cloud in the sky?
|
a large body of water
|
a cloud in the sky
| null | null |
[
{
"bbox_3d": [
35.9,
6.3,
174.4
],
"label": "a black boat"
},
{
"bbox_3d": [
6,
2.7,
60.9
],
"label": "a large body of water"
},
{
"bbox_3d": [
0.1,
33,
233.4
],
"label": "a cloud in the sky"
}
] |
[
{
"front_dir": [
-0.1,
-0.1,
-1
],
"label": "a black boat",
"left_dir": [
-1,
0,
0.1
]
}
] |
A
|
To solve this problem, we first detect the 3D location of a black boat, a large body of water, and a cloud in the sky. Then we compute the cosine similarities between the front direction of a black boat and the vectors from a black boat to the other two objects. We can estimate the angles from the cosine similarities, and the object with a smaller angle is the object that a black boat is facing towards. The 3D location of a black boat is (35.9, 6.3, 174.4). The 3D location of a large body of water is (6.0, 2.7, 60.9). The 3D location of a cloud in the sky is (0.1, 33.0, 233.4). The front direction of a black boat is (-0.1, -0.1, -1.0). First we consider if a black boat is facing towards the a large body of water. The vector from a black boat to a large body of water is (-29.9, -3.6, -113.5). The cosine similarity between the front direction and the vector is 0.99, corresponding to an angle of 8.65 degrees. First we consider if a black boat is facing towards the a cloud in the sky. The vector from a black boat to a cloud in the sky is (-35.8, 26.8, 59.0). The cosine similarity between the front direction and the vector is -0.76, corresponding to an angle of 139.83 degrees. We find that the angle between the front direction and a large body of water is smaller. Therefore, the final answer is A. a large body of water.
|
A. a large body of water.
|
multi_object_facing
|
007390fcb5346eff.jpg
|
0073d756da528f00_b653
|
Consider the real-world 3D locations of the objects. Which object has a higher location?
|
a boat on a lake
|
a yellow float with a face on it
| null | null |
[
{
"bbox_3d": [
-2.1,
1.7,
30.6
],
"label": "a boat on a lake"
},
{
"bbox_3d": [
0.2,
4.6,
47.3
],
"label": "a yellow float with a face on it"
}
] |
[] |
A
|
To solve this problem, we first estimate the 3D heights of the two objects. The object with a larger height value is at a higher location. The object with a smaller height value is at a lower location. The 3D height of a boat on a lake is 7.4. The 3D height of a yellow float with a face on it is 5.4. The 3D height of a boat on a lake is larger, meaning that the location of a boat on a lake is higher. Therefore, the answer is A. a yellow float with a face on it.
|
A. a yellow float with a face on it.
|
height_higher
|
0073d756da528f00.jpg
|
0073f1693027f10b_ef3a
|
Consider the real-world 3D locations of the objects. Which object has a higher location?
|
a girl wearing a pink and yellow bikini
|
a woman wearing a pink and blue bikini
| null | null |
[
{
"bbox_3d": [
-0.1,
0.5,
2.7
],
"label": "a girl wearing a pink and yellow bikini"
},
{
"bbox_3d": [
-0.1,
0.3,
2.7
],
"label": "a woman wearing a pink and blue bikini"
}
] |
[] |
A
|
To solve this problem, we first estimate the 3D heights of the two objects. The object with a larger height value is at a higher location. The object with a smaller height value is at a lower location. The 3D height of a girl wearing a pink and yellow bikini is 1.4. The 3D height of a woman wearing a pink and blue bikini is 0.4. The 3D height of a girl wearing a pink and yellow bikini is larger, meaning that the location of a girl wearing a pink and yellow bikini is higher. Therefore, the answer is A. a woman wearing a pink and blue bikini.
|
A. a woman wearing a pink and blue bikini.
|
height_higher
|
0073f1693027f10b.jpg
|
00745e321470b42c_d6b8
|
Consider the real-world 3D location of the objects. Which object is further away from the camera?
|
a black bicycle with a black seat
|
a young girl wearing a pink helmet
| null | null |
[
{
"bbox_3d": [
-1.3,
0.5,
3.7
],
"label": "a black bicycle with a black seat"
},
{
"bbox_3d": [
-1.1,
0.6,
5
],
"label": "a young girl wearing a pink helmet"
}
] |
[] |
B
|
To solve this problem, we first estimate the 3D locations of a black bicycle with a black seat and a young girl wearing a pink helmet. Then we estimate the L2 distances from the camera to the two objects. The object with a smaller L2 distance is the one that is closer to the camera. The 3D location of a black bicycle with a black seat is (-1.3, 0.5, 3.7). The 3D location of a young girl wearing a pink helmet is (-1.1, 0.6, 5.0). The L2 distance from the camera to a black bicycle with a black seat is 3.99. The L2 distance from the camera to a young girl wearing a pink helmet is 5.16. The distance to a young girl wearing a pink helmet is larger. Therefore, the answer is B. a young girl wearing a pink helmet.
|
B. a young girl wearing a pink helmet.
|
location_closer_to_camera
|
00745e321470b42c.jpg
|
00745e321470b42c_7b56
|
Consider the real-world 3D locations and orientations of the objects. Which object is a bicycle with a purple frame facing towards, a boy wearing a blue helmet or the a bicycle with a pink frame?
|
a boy wearing a blue helmet
|
a bicycle with a pink frame
| null | null |
[
{
"bbox_3d": [
0.5,
0.5,
5.8
],
"label": "a bicycle with a purple frame"
},
{
"bbox_3d": [
2,
1,
5.5
],
"label": "a boy wearing a blue helmet"
},
{
"bbox_3d": [
1,
0.4,
3.9
],
"label": "a bicycle with a pink frame"
}
] |
[
{
"front_dir": [
0.4,
-0.4,
-0.8
],
"label": "a bicycle with a purple frame",
"left_dir": [
-0.9,
-0.2,
-0.4
]
},
{
"front_dir": [
0.3,
-0.3,
-0.9
],
"label": "a bicycle with a pink frame",
"left_dir": [
-0.9,
-0.2,
-0.3
]
}
] |
B
|
To solve this problem, we first detect the 3D location of a bicycle with a purple frame, a boy wearing a blue helmet, and a bicycle with a pink frame. Then we compute the cosine similarities between the front direction of a bicycle with a purple frame and the vectors from a bicycle with a purple frame to the other two objects. We can estimate the angles from the cosine similarities, and the object with a smaller angle is the object that a bicycle with a purple frame is facing towards. The 3D location of a bicycle with a purple frame is (0.5, 0.5, 5.8). The 3D location of a boy wearing a blue helmet is (2.0, 1.0, 5.5). The 3D location of a bicycle with a pink frame is (1.0, 0.4, 3.9). The front direction of a bicycle with a purple frame is (0.4, -0.4, -0.8). First we consider if a bicycle with a purple frame is facing towards the a boy wearing a blue helmet. The vector from a bicycle with a purple frame to a boy wearing a blue helmet is (1.5, 0.5, -0.3). The cosine similarity between the front direction and the vector is 0.46, corresponding to an angle of 62.37 degrees. First we consider if a bicycle with a purple frame is facing towards the a bicycle with a pink frame. The vector from a bicycle with a purple frame to a bicycle with a pink frame is (0.5, -0.1, -1.9). The cosine similarity between the front direction and the vector is 0.93, corresponding to an angle of 21.65 degrees. We find that the angle between the front direction and a bicycle with a pink frame is smaller. Therefore, the final answer is B. a bicycle with a pink frame.
|
B. a bicycle with a pink frame.
|
multi_object_facing
|
00745e321470b42c.jpg
|
00750e1c4ed31816_6118
|
Consider the real-world 3D locations and orientations of the objects. Which side of a black and green kayak in the water is facing the camera?
|
front
|
left
|
back
|
right
|
[
{
"bbox_3d": [
0,
0.4,
14.2
],
"label": "a black and green kayak in the water"
}
] |
[
{
"front_dir": [
-1,
0.1,
-0.1
],
"label": "a black and green kayak in the water",
"left_dir": [
-0.1,
-0.2,
1
]
}
] |
D
|
To solve this problem, we first estimate the 3D location of a black and green kayak in the water. Then we obtain the vector pointing from the object to the camera. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a black and green kayak in the water, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a black and green kayak in the water that is facing the camera corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The location of a black and green kayak in the water is (0.0, 0.4, 14.2). The vector from a black and green kayak in the water to camera is hence (-0.0, -0.4, -14.2). The left direction of a black and green kayak in the water is (-0.1, -0.2, 1.0). The cosine similarity between the vector pointing to camera and the left direction is -0.98, corresponding to an angle of 168.02 degrees. Thus the angle between the vector pointing to camera and the right direction is 11.98 degrees. The front direction of a black and green kayak in the water is (-1.0, 0.1, -0.1). The cosine similarity between the vector pointing to camera and the front direction is 0.05, corresponding to an angle of 87.16 degrees. Thus the angle between the vector pointing to camera and the back direction is 92.84 degrees. Among the four directions, the smallest angle is the right direction, with an angle of 11.98 degrees. Thus the right side of a black and green kayak in the water is facing the camera. Therefore, the final answer is D. right.
|
D. right.
|
orientation_viewpoint
|
00750e1c4ed31816.jpg
|
00758c10b2da2954_d1ca
|
Consider the real-world 3D locations of the objects. Which object has a lower location?
|
a girl holding two phones
|
a black smartphone with a green leaf on the screen
| null | null |
[
{
"bbox_3d": [
0.3,
0.3,
1
],
"label": "a girl holding two phones"
},
{
"bbox_3d": [
-0.1,
0.2,
0.7
],
"label": "a black smartphone with a green leaf on the screen"
}
] |
[] |
B
|
To solve this problem, we first estimate the 3D heights of the two objects. The object with a larger height value is at a higher location. The object with a smaller height value is at a lower location. The 3D height of a girl holding two phones is 1.1. The 3D height of a black smartphone with a green leaf on the screen is 0.4. The 3D height of a girl holding two phones is larger, meaning that the location of a girl holding two phones is higher. In other words, the location of a black smartphone with a green leaf on the screen is lower. Therefore, the answer is B. a girl holding two phones.
|
B. a girl holding two phones
|
height_higher
|
00758c10b2da2954.jpg
|
007692b77a74a459_e040
|
Consider the real-world 3D orientations of the objects. Are a bulletin board with many pictures and a green chalkboard facing same or similar directions, or very different directions?
|
same or similar directions
|
very different directions
| null | null |
[
{
"bbox_3d": [
-0.3,
1.8,
5.1
],
"label": "a bulletin board with many pictures"
},
{
"bbox_3d": [
-0.6,
0.9,
3.7
],
"label": "a green chalkboard"
}
] |
[
{
"front_dir": [
0.9,
-0.2,
-0.5
],
"label": "a bulletin board with many pictures",
"left_dir": [
-0.5,
-0.2,
-0.9
]
},
{
"front_dir": [
0.6,
-0.1,
-0.8
],
"label": "a green chalkboard",
"left_dir": [
-0.8,
0.1,
-0.6
]
}
] |
A
|
To solve this problem, we first detect the front directions of a bulletin board with many pictures and a green chalkboard. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is small, then the two objects are facing same or similar directions. Otherwise, the two objects are facing very different directions. The front direction of a bulletin board with many pictures is (0.9, -0.2, -0.5). The front direction of a green chalkboard is (0.6, -0.1, -0.8). The cosine similarity between the two front directions is 0.92, corresponding to an angle of 23.48. The angle is small, meaning that the two objects are facing same or similar directions. Therefore, the final answer is A. same or similar directions.
|
A. same or similar directions.
|
multi_object_same_direction
|
007692b77a74a459.jpg
|
00784ef6ad094836_3187
|
Consider the real-world 3D orientations of the objects. What is the relationship between the orientations of a black computer monitor and a white laptop computer, parallel of perpendicular to each other?
|
parallel
|
perpendicular
| null | null |
[
{
"bbox_3d": [
-0.2,
1,
8.3
],
"label": "a black computer monitor"
},
{
"bbox_3d": [
1.5,
1.2,
9
],
"label": "a white laptop computer"
}
] |
[
{
"front_dir": [
0.1,
0.1,
-1
],
"label": "a black computer monitor",
"left_dir": [
-1,
0.1,
-0.1
]
},
{
"front_dir": [
-0.1,
0,
-1
],
"label": "a white laptop computer",
"left_dir": [
-1,
0.1,
0.1
]
}
] |
A
|
To solve this problem, we first detect the front directions of a black computer monitor and a white laptop computer. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is closer to 0 or 180, than to 90 degrees, then the two objects are parallel to each other. Otherwise, the two objects are perpendicular to each other. The front direction of a black computer monitor is (0.1, 0.1, -1.0). The front direction of a white laptop computer is (-0.1, -0.0, -1.0). The cosine similarity between the two front directions is 0.96, corresponding to an angle of 15.58. The angle is closer to 0 or 180, than to 90 degrees, meaning that the two objects are parallel to each other. Therefore, the final answer is A. parallel.
|
A. parallel.
|
multi_object_parallel
|
00784ef6ad094836.jpg
|
00785b842dbe3f27_01d6
|
Consider the real-world 3D locations and orientations of the objects. Which side of a white plane with a blue stripe is facing the camera?
|
front
|
left
|
back
|
right
|
[
{
"bbox_3d": [
0.4,
2.8,
6.9
],
"label": "a white plane with a blue stripe"
}
] |
[
{
"front_dir": [
0.9,
-0.4,
-0.2
],
"label": "a white plane with a blue stripe",
"left_dir": [
-0.4,
-0.3,
-0.9
]
}
] |
B
|
To solve this problem, we first estimate the 3D location of a white plane with a blue stripe. Then we obtain the vector pointing from the object to the camera. Now we compute the angles between the vector and the left, right, front, back directions. We first compute the left direction of a white plane with a blue stripe, which leads to the angle between left direction and the vector. Immediately we also get the angle between the right direction and the vector. Similarly we compute the angles between the vector and front, back directions. The side of a white plane with a blue stripe that is facing the camera corresponds to the direction with the smallest angle. Based on the four estimated angles, we can predict the final answer. The location of a white plane with a blue stripe is (0.4, 2.8, 6.9). The vector from a white plane with a blue stripe to camera is hence (-0.4, -2.8, -6.9). The left direction of a white plane with a blue stripe is (-0.4, -0.3, -0.9). The cosine similarity between the vector pointing to camera and the left direction is 0.95, corresponding to an angle of 18.30 degrees. Thus the angle between the vector pointing to camera and the right direction is 161.70 degrees. The front direction of a white plane with a blue stripe is (0.9, -0.4, -0.2). The cosine similarity between the vector pointing to camera and the front direction is 0.31, corresponding to an angle of 71.85 degrees. Thus the angle between the vector pointing to camera and the back direction is 108.15 degrees. Among the four directions, the smallest angle is the left direction, with an angle of 18.30 degrees. Thus the left side of a white plane with a blue stripe is facing the camera. Therefore, the final answer is B. left.
|
B. left.
|
orientation_viewpoint
|
00785b842dbe3f27.jpg
|
0078c9a3a0115218_7318
|
Consider the real-world 3D locations and orientations of the objects. If I stand at a black suitcase with a yellow sticker on it's position facing where it is facing, is a wheel with a black cover on the left or right of me?
|
on the left
|
on the right
| null | null |
[
{
"bbox_3d": [
-0.3,
0.2,
0.8
],
"label": "a wheel with a black cover"
},
{
"bbox_3d": [
0.2,
0.6,
1
],
"label": "a black suitcase with a yellow sticker on it"
}
] |
[
{
"front_dir": [
-0.1,
0.2,
-1
],
"label": "a black suitcase with a yellow sticker on it",
"left_dir": [
-1,
0,
0.1
]
}
] |
A
|
To solve this problem, we first determine the 3D locations of a wheel with a black cover and a black suitcase with a yellow sticker on it. Then we estimate the vector pointing from a black suitcase with a yellow sticker on it to a wheel with a black cover, as well as the left direction of a black suitcase with a yellow sticker on it. Next we compute the cosine similarities between the vector and the left direction, which leads to the angle between the vector and the fron direction. If the angle is smaller than 90 degrees, then a wheel with a black cover is on the left of a black suitcase with a yellow sticker on it. Otherwise, a wheel with a black cover is behind a black suitcase with a yellow sticker on it. The 3D location of a wheel with a black cover is (-0.3, 0.2, 0.8). The 3D location of a black suitcase with a yellow sticker on it is (0.2, 0.6, 1.0). The vector from a black suitcase with a yellow sticker on it to a wheel with a black cover is hence (-0.5, -0.4, -0.2). The left direction of a black suitcase with a yellow sticker on it is (-1.0, -0.0, 0.1). The cosine similarity between the vector and the left direction is 0.76, corresponding to an angle of 40.97 degrees. The angle is smaller than 90 degrees, meaning that a wheel with a black cover is on the left of a black suitcase with a yellow sticker on it. Therefore, the final answer is A. on the left.
|
A. on the left.
|
orientation_on_the_left
|
0078c9a3a0115218.jpg
|
007971cdc01ea821_1ca8
|
Consider the real-world 3D locations and orientations of the objects. Which object is a small car facing towards, a large tusk with a yellow tip or the a white car?
|
a large tusk with a yellow tip
|
a white car
| null | null |
[
{
"bbox_3d": [
14.7,
1.6,
27.3
],
"label": "a small car"
},
{
"bbox_3d": [
0.7,
2.1,
3.9
],
"label": "a large tusk with a yellow tip"
},
{
"bbox_3d": [
-2,
2.3,
26.3
],
"label": "a white car"
}
] |
[
{
"front_dir": [
-0.4,
-0.1,
-0.9
],
"label": "a small car",
"left_dir": [
-0.9,
0.1,
0.4
]
},
{
"front_dir": [
0.4,
-0.1,
-0.9
],
"label": "a white car",
"left_dir": [
-0.9,
0,
-0.5
]
}
] |
A
|
To solve this problem, we first detect the 3D location of a small car, a large tusk with a yellow tip, and a white car. Then we compute the cosine similarities between the front direction of a small car and the vectors from a small car to the other two objects. We can estimate the angles from the cosine similarities, and the object with a smaller angle is the object that a small car is facing towards. The 3D location of a small car is (14.7, 1.6, 27.3). The 3D location of a large tusk with a yellow tip is (0.7, 2.1, 3.9). The 3D location of a white car is (-2.0, 2.3, 26.3). The front direction of a small car is (-0.4, -0.1, -0.9). First we consider if a small car is facing towards the a large tusk with a yellow tip. The vector from a small car to a large tusk with a yellow tip is (-13.9, 0.5, -23.4). The cosine similarity between the front direction and the vector is 0.99, corresponding to an angle of 9.04 degrees. First we consider if a small car is facing towards the a white car. The vector from a small car to a white car is (-16.7, 0.7, -0.9). The cosine similarity between the front direction and the vector is 0.45, corresponding to an angle of 62.99 degrees. We find that the angle between the front direction and a large tusk with a yellow tip is smaller. Therefore, the final answer is A. a large tusk with a yellow tip.
|
A. a large tusk with a yellow tip.
|
multi_object_facing
|
007971cdc01ea821.jpg
|
007a74030db075a4_e9ad
|
Consider the real-world 3D locations and orientations of the objects. Which object is a black chair with a man sitting in it facing towards, a black laptop on a desk or the a computer monitor with a blue base?
|
a black laptop on a desk
|
a computer monitor with a blue base
| null | null |
[
{
"bbox_3d": [
0.2,
0.6,
2.1
],
"label": "a black chair with a man sitting in it"
},
{
"bbox_3d": [
-0.2,
0.8,
1.6
],
"label": "a black laptop on a desk"
},
{
"bbox_3d": [
-0.2,
1.9,
2.6
],
"label": "a computer monitor with a blue base"
}
] |
[
{
"front_dir": [
-0.9,
0.1,
-0.4
],
"label": "a black chair with a man sitting in it",
"left_dir": [
-0.4,
-0.1,
0.9
]
},
{
"front_dir": [
-0.8,
0.1,
-0.6
],
"label": "a black laptop on a desk",
"left_dir": [
-0.6,
-0.1,
0.8
]
},
{
"front_dir": [
0.6,
-0.3,
-0.7
],
"label": "a computer monitor with a blue base",
"left_dir": [
-0.8,
-0.1,
-0.6
]
}
] |
A
|
To solve this problem, we first detect the 3D location of a black chair with a man sitting in it, a black laptop on a desk, and a computer monitor with a blue base. Then we compute the cosine similarities between the front direction of a black chair with a man sitting in it and the vectors from a black chair with a man sitting in it to the other two objects. We can estimate the angles from the cosine similarities, and the object with a smaller angle is the object that a black chair with a man sitting in it is facing towards. The 3D location of a black chair with a man sitting in it is (0.2, 0.6, 2.1). The 3D location of a black laptop on a desk is (-0.2, 0.8, 1.6). The 3D location of a computer monitor with a blue base is (-0.2, 1.9, 2.6). The front direction of a black chair with a man sitting in it is (-0.9, 0.1, -0.4). First we consider if a black chair with a man sitting in it is facing towards the a black laptop on a desk. The vector from a black chair with a man sitting in it to a black laptop on a desk is (-0.4, 0.2, -0.5). The cosine similarity between the front direction and the vector is 0.91, corresponding to an angle of 24.99 degrees. First we consider if a black chair with a man sitting in it is facing towards the a computer monitor with a blue base. The vector from a black chair with a man sitting in it to a computer monitor with a blue base is (-0.4, 1.2, 0.5). The cosine similarity between the front direction and the vector is 0.22, corresponding to an angle of 77.39 degrees. We find that the angle between the front direction and a black laptop on a desk is smaller. Therefore, the final answer is A. a black laptop on a desk.
|
A. a black laptop on a desk.
|
multi_object_facing
|
007a74030db075a4.jpg
|
007b42b566c8950a_41fc
|
Consider the real-world 3D locations of the objects. Are the a white license plate with black letters and numbers and the a green ivy plant next to each other or far away from each other?
|
next to each other
|
far away from each other
| null | null |
[
{
"bbox_3d": [
-3.1,
0.6,
8.3
],
"label": "a white license plate with black letters and numbers"
},
{
"bbox_3d": [
-0.4,
1.5,
7.2
],
"label": "a green ivy plant"
}
] |
[] |
B
|
To solve this problem, we first estimate the 3D locations of a white license plate with black letters and numbers and a green ivy plant. Then we can compute the L2 distance between the two objects. Next we estimate the rough sizes of the two objects. If the distance between the two objects is smaller or roughly the same as the object sizes, then the two objects are next to each other. Otherwise, the two objects are far away from each other. The 3D location of a white license plate with black letters and numbers is (-3.1, 0.6, 8.3). The 3D location of a green ivy plant is (-0.4, 1.5, 7.2). The L2 distance between the two objects is 3.06. The size of the a white license plate with black letters and numbers is roughly 0.54. The size of the a green ivy plant is roughly 2.11. The distance between the two objects is much bigger than the object sizes, meaning that they are far away from each other. Therefore, the answer is B. far away from each other.
|
B. far away from each other.
|
location_next_to
|
007b42b566c8950a.jpg
|
007b77fbf7c6c7c2_dc53
|
Consider the real-world 3D orientations of the objects. What is the relationship between the orientations of a white keyboard with a silver and white color scheme and a computer monitor with a keyboard in front of it, parallel of perpendicular to each other?
|
parallel
|
perpendicular
| null | null |
[
{
"bbox_3d": [
0,
0.3,
0.9
],
"label": "a white keyboard with a silver and white color scheme"
},
{
"bbox_3d": [
0.3,
0.7,
1
],
"label": "a computer monitor with a keyboard in front of it"
}
] |
[
{
"front_dir": [
0.3,
-0.1,
-0.9
],
"label": "a white keyboard with a silver and white color scheme",
"left_dir": [
-0.9,
0,
-0.3
]
},
{
"front_dir": [
0.1,
0,
-1
],
"label": "a computer monitor with a keyboard in front of it",
"left_dir": [
-1,
0,
-0.1
]
}
] |
A
|
To solve this problem, we first detect the front directions of a white keyboard with a silver and white color scheme and a computer monitor with a keyboard in front of it. Then we compute the cosine similarities between the two front directions, and the angle between them. If the angle between the two front directions is closer to 0 or 180, than to 90 degrees, then the two objects are parallel to each other. Otherwise, the two objects are perpendicular to each other. The front direction of a white keyboard with a silver and white color scheme is (0.3, -0.1, -0.9). The front direction of a computer monitor with a keyboard in front of it is (0.1, 0.0, -1.0). The cosine similarity between the two front directions is 0.96, corresponding to an angle of 16.86. The angle is closer to 0 or 180, than to 90 degrees, meaning that the two objects are parallel to each other. Therefore, the final answer is A. parallel.
|
A. parallel.
|
multi_object_parallel
|
007b77fbf7c6c7c2.jpg
|
007b8e65878e55c6_34e0
|
Consider the real-world 3D locations and orientations of the objects. Which object is a yellow school bus parked in a grassy area facing towards, a small tree with green leaves or the a tall green tree?
|
a small tree with green leaves
|
a tall green tree
| null | null |
[
{
"bbox_3d": [
14.2,
3.9,
63.2
],
"label": "a yellow school bus parked in a grassy area"
},
{
"bbox_3d": [
11,
4.6,
50.9
],
"label": "a small tree with green leaves"
},
{
"bbox_3d": [
9.1,
9.7,
78
],
"label": "a tall green tree"
}
] |
[
{
"front_dir": [
-0.1,
0.1,
-1
],
"label": "a yellow school bus parked in a grassy area",
"left_dir": [
-1,
0,
0.1
]
}
] |
A
|
To solve this problem, we first detect the 3D location of a yellow school bus parked in a grassy area, a small tree with green leaves, and a tall green tree. Then we compute the cosine similarities between the front direction of a yellow school bus parked in a grassy area and the vectors from a yellow school bus parked in a grassy area to the other two objects. We can estimate the angles from the cosine similarities, and the object with a smaller angle is the object that a yellow school bus parked in a grassy area is facing towards. The 3D location of a yellow school bus parked in a grassy area is (14.2, 3.9, 63.2). The 3D location of a small tree with green leaves is (11.0, 4.6, 50.9). The 3D location of a tall green tree is (9.1, 9.7, 78.0). The front direction of a yellow school bus parked in a grassy area is (-0.1, 0.1, -1.0). First we consider if a yellow school bus parked in a grassy area is facing towards the a small tree with green leaves. The vector from a yellow school bus parked in a grassy area to a small tree with green leaves is (-3.2, 0.7, -12.2). The cosine similarity between the front direction and the vector is 0.99, corresponding to an angle of 6.39 degrees. First we consider if a yellow school bus parked in a grassy area is facing towards the a tall green tree. The vector from a yellow school bus parked in a grassy area to a tall green tree is (-5.1, 5.8, 14.8). The cosine similarity between the front direction and the vector is -0.82, corresponding to an angle of 144.66 degrees. We find that the angle between the front direction and a small tree with green leaves is smaller. Therefore, the final answer is A. a small tree with green leaves.
|
A. a small tree with green leaves.
|
multi_object_facing
|
007b8e65878e55c6.jpg
|
007bae2adf43dd31_0a8d
|
Consider the real-world 3D locations and orientations of the objects. If I stand at a blue tray on a desk's position facing where it is facing, is a gray chair with wheels on the left or right of me?
|
on the left
|
on the right
| null | null |
[
{
"bbox_3d": [
1.3,
0.8,
2.8
],
"label": "a gray chair with wheels"
},
{
"bbox_3d": [
0.2,
1,
1.8
],
"label": "a blue tray on a desk"
}
] |
[
{
"front_dir": [
-0.8,
-0.1,
-0.6
],
"label": "a gray chair with wheels",
"left_dir": [
-0.6,
0,
0.8
]
},
{
"front_dir": [
0.7,
0,
-0.7
],
"label": "a blue tray on a desk",
"left_dir": [
-0.7,
0.1,
-0.7
]
}
] |
B
|
To solve this problem, we first determine the 3D locations of a gray chair with wheels and a blue tray on a desk. Then we estimate the vector pointing from a blue tray on a desk to a gray chair with wheels, as well as the left direction of a blue tray on a desk. Next we compute the cosine similarities between the vector and the left direction, which leads to the angle between the vector and the fron direction. If the angle is smaller than 90 degrees, then a gray chair with wheels is on the left of a blue tray on a desk. Otherwise, a gray chair with wheels is behind a blue tray on a desk. The 3D location of a gray chair with wheels is (1.3, 0.8, 2.8). The 3D location of a blue tray on a desk is (0.2, 1.0, 1.8). The vector from a blue tray on a desk to a gray chair with wheels is hence (1.1, -0.3, 1.0). The left direction of a blue tray on a desk is (-0.7, 0.1, -0.7). The cosine similarity between the vector and the left direction is -0.99, corresponding to an angle of 173.86 degrees. The angle is smaller than 90 degrees, meaning that a gray chair with wheels is on the right of a blue tray on a desk. Therefore, the final answer is B. on the right.
|
B. on the right.
|
orientation_on_the_left
|
007bae2adf43dd31.jpg
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.