# YOLOv9 Card Detector

This model is a fine-tuned version of YOLOv9c trained to detect playing cards in images. It has been trained on the Set Cards dataset from Roboflow.

## Model Details

- **Base Model**: YOLOv9c
- **Task**: Object Detection
- **Target Class**: Cards
- **Training Dataset**: [Set Cards Dataset](https://universe.roboflow.com/tel-aviv/set_cards/dataset/1)
- **Image Size**: 512x512
- **Accuracy Metrics**: Evaluated at confidence threshold of 0.5

## Usage

```python
from transformers import AutoImageProcessor, AutoModelForObjectDetection
import torch
from PIL import Image
import requests

# Load model and processor
processor = AutoImageProcessor.from_pretrained("YOUR_USERNAME/yolov9-card-detector")
model = AutoModelForObjectDetection.from_pretrained("YOUR_USERNAME/yolov9-card-detector")

# Load image
image_url = "https://example.com/path/to/card_image.jpg"
image = Image.open(requests.get(image_url, stream=True).raw)

# Prepare image for the model
inputs = processor(images=image, return_tensors="pt")

# Make prediction
with torch.no_grad():
    outputs = model(**inputs)

# Process results
results = processor.post_process_object_detection(
    outputs,
    threshold=0.5,
    target_sizes=[(image.height, image.width)]
)[0]

# Display results
for score, label, box in zip(results["scores"], results["labels"], results["boxes"]):
    box = [round(i, 2) for i in box.tolist()]
    print(
        f"Detected {model.config.id2label[label.item()]} with confidence "
        f"{round(score.item(), 3)} at location {box}"
    )
```

## Training

This model was fine-tuned from YOLOv9c using the Ultralytics framework. It was trained for 30 epochs with an image size of 512x512.

## License

This model is licensed under CC BY 4.0, following the dataset's licensing terms.

## Limitations

- The model is specifically trained to detect playing cards and may not perform well on other objects
- Performance may vary based on lighting conditions, card orientation, and image quality
- Best results are achieved with images similar to those in the training dataset