Upload MedVQA_test.py
Browse files- MedVQA_test.py +58 -0
MedVQA_test.py
ADDED
@@ -0,0 +1,58 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import csv
|
3 |
+
import datasets
|
4 |
+
from PIL import Image
|
5 |
+
import pandas as pd
|
6 |
+
|
7 |
+
|
8 |
+
class MyDataset(datasets.GeneratorBasedBuilder):
|
9 |
+
def _info(self):
|
10 |
+
return datasets.DatasetInfo(
|
11 |
+
description="Dataset containing medical images with associated questions and answers.",
|
12 |
+
features=datasets.Features({
|
13 |
+
"image": datasets.Image(),
|
14 |
+
"source": datasets.Value("string"),
|
15 |
+
"question": datasets.Value("string"),
|
16 |
+
"answer": datasets.Value("string"),
|
17 |
+
"img_id": datasets.Value("string"), }),
|
18 |
+
supervised_keys=None,
|
19 |
+
)
|
20 |
+
|
21 |
+
def _split_generators(self, dl_manager):
|
22 |
+
df = pd.read_csv(os.path.join(self.config.data_dir, "gt.csv"),
|
23 |
+
delimiter=";")
|
24 |
+
return [
|
25 |
+
datasets.SplitGenerator(
|
26 |
+
name=datasets.Split.VALIDATION,
|
27 |
+
gen_kwargs={"df": df,
|
28 |
+
"images_dir": os.path.join(self.config.data_dir, "data")}
|
29 |
+
)
|
30 |
+
]
|
31 |
+
|
32 |
+
def _generate_examples(self, df, images_dir):
|
33 |
+
# iterate df
|
34 |
+
for idx, row in df.iterrows():
|
35 |
+
img_id = row.get("img_id")
|
36 |
+
if not img_id:
|
37 |
+
continue # Skip rows without a valid image identifier
|
38 |
+
image_path = os.path.join(images_dir, f"{img_id}")
|
39 |
+
if not os.path.exists(image_path):
|
40 |
+
continue # Skip if image file does not exist
|
41 |
+
yield idx, {
|
42 |
+
"image": image_path,
|
43 |
+
"source": row["source"],
|
44 |
+
"question": row["question"],
|
45 |
+
"answer": row["answer"],
|
46 |
+
"img_id": img_id.split(".")[0],
|
47 |
+
}
|
48 |
+
|
49 |
+
# MedVQA_test_cli.py
|
50 |
+
# from datasets import load_dataset
|
51 |
+
|
52 |
+
# dataset = load_dataset(
|
53 |
+
# "/Users/sgautam/Documents/MedVQA/MedVQA_test.py", data_dir="/Users/sgautam/Downloads/kvasir_vqa_test/", trust_remote_code=True)
|
54 |
+
# dataset.push_to_hub("SimulaMet/Kvasir-VQA-test", private=False)
|
55 |
+
# breakpoint()
|
56 |
+
|
57 |
+
# then
|
58 |
+
# python MedVQA_test_cli.py
|