Chest CT Scan Classifier (Streamlit)

A simple, good-looking Streamlit web app that loads a ConvNeXt-Large PyTorch model from CTScan_ConvNeXtLarge.pth and predicts the class probabilities for an uploaded chest CT image.

Features

• Upload PNG/JPG chest CT images
• Auto-preprocessing for ConvNeXt (resize 224, ImageNet normalization)
• Top-K predictions with confidence bar chart
• Generic labels generated automatically (Class 0..N)

(Optional) Create & activate virtual environment

python -m venv .venv; ..venv\Scripts\Activate.ps1

Install dependencies

pip install -r requirements.txt

Run the app

streamlit run app.py

## Model
- Place the trained checkpoint at `CTScan_ConvNeXtLarge.pth` (already present).
- The app tries to infer the number of classes from the checkpoint's classifier weight shape.

## Notes
- This app is not a medical device; predictions are probabilistic.
- If your model uses a different preprocessing pipeline, adjust the `PREPROCESS` transform in `app.py`.