Pneumothorax Classification

Overview

This repository contains code and resources for classifying pneumothorax conditions from medical images. The project leverages deep learning techniques to analyze and predict the presence of pneumothorax in chest X-ray images.

Prerequisites

• Docker
• Internet connection for downloading model weights

Dataset

• train and test image datasets abhishek/siim-png-images
• Mapping csv for training and testing chogerlate/siim-train-dataframe

Weights Downloading

Before building the Docker image, you need to download the pre-trained model weights. Follow these steps:

1. Create a directory named weights in the root of the repository:

   mkdir weights

2. Install gdown to download model weight from Google Drive

   pip install gdown

3. Download the weights file from the provided link and place it in the weights directory.

   gdown https://drive.google.com/uc?id=<file-id>

   Note that the URL might change in the future; if it is, you can DM me.
4. Copy downloaded weight into weights directory.

Building the Docker Image

To build the Docker image, navigate to the root of the repository and run the following command:

docker build -t pneumothorax-classification .

Running the Docker Container

Once the Docker image is built, you can run the container using the following command:

docker run -p 80:80 pneumothorax-classification

This will start the application, and it will be accessible at http://localhost:80.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Acknowledgements

• Dataset Source
• Pre-trained Model --> DM ME!

For more details, refer to the documentation or contact the project maintainers.