jayanthapoojary1989/rsna-pneumonia-faster-rcnn

---

    title: RSNA Pneumonia Detection Faster R-CNN
    tags:
    - object-detection
    - medical
    - pneumonia
    - faster-rcnn
    - pytorch
    library_name: torchvision
    ---

    # RSNA Pneumonia Detection Model (Faster R-CNN ResNet50-FPN)

    This repository contains a Faster R-CNN ResNet50-FPN model trained for detecting Pneumonia (Lung Opacity) from chest X-ray images, based on the RSNA Pneumonia Detection Challenge dataset.

    ## Model Details
    - **Architecture**: Faster R-CNN ResNet50-FPN
    - **Task**: Object Detection
    - **Classes**: `background`, `pneumonia` (2 classes total)
    - **Input Image Size**: 512x512
    - **Training Data**: Subset of RSNA Pneumonia Detection Challenge dataset.

    ## How to Use
    You can load this model using PyTorch and Torchvision:

    ```python
    import torch
    import torchvision
    from torchvision.models.detection.faster_rcnn import FastRCNNPredictor

    # Define your model architecture
    def get_model(num_classes):
        model = torchvision.models.detection.fasterrcnn_resnet50_fpn(
            weights=torchvision.models.detection.FasterRCNN_ResNet50_FPN_Weights.DEFAULT
        )
        in_features = model.roi_heads.box_predictor.cls_score.in_features
        model.roi_heads.box_predictor = FastRCNNPredictor(in_features, num_classes)
        return model

    # Load the model directly from the Hugging Face Hub
    # Ensure you have the 'accelerate' library installed for download progress
    # pip install accelerate

    # Create a dummy model instance to load state_dict into
    num_classes = 2 # 2 for background and pneumonia
    model = get_model(num_classes)

    # Load the state_dict
    # The model file will be downloaded by the HfApi internally
    from huggingface_hub import hf_hub_download
    model_path_in_hub = hf_hub_download(repo_id="jayanthapoojary1989/rsna-pneumonia-faster-rcnn", filename="faster_rcnn_pneumonia_model.pth")
    model.load_state_dict(torch.load(model_path_in_hub, map_location='cpu')) # Use 'cpu' for loading then move to device
    model.eval() # Set to evaluation mode

    # Example inference (assuming 'image' is a preprocessed tensor suitable for the model)
    # You would load and preprocess your image here (e.g., PIL Image -> ToTensor)
    # image = your_transform(PIL.Image.open("path/to/image.jpg")).unsqueeze(0) # Add batch dim
    # device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
    # model.to(device)
    # image = image.to(device)
    # with torch.no_grad():
    #     predictions = model(image)

    # print(predictions)
    Disclaimer
    This model is provided for research and educational purposes. Use in clinical settings requires rigorous validation, regulatory approval, and expert medical supervision.