This repository aims to solve the problem of Pneumonia Detection via AI which is the Kaggle RSNA Challenge.
We performed preprocessing, training, visualization (GradCam) and detection using the given dataset in Kaggle RSNA.

Requires Python 3.6 and keras 2.2.4, tensorflow-gpu 1.14.0, implemented by Cumtom Retinanet.

$ pip install -r requirements.txt

1. First, edit the SETTINGS.json file to suit your personal environment.
2. Save Kaggle RSNA DataSet in ./data/ folder.
3. Run prepare_data.sh.

$ sh prepare_data.sh

Modify Retinanet.py or train.sh.

$ sh train.sh

or

$ python ./src/train/Retinanet/Retinanet.py

Modify Eval_Retinanet.py or evaluate.sh.

$ sh evaluate.sh

or

$ python ./src/eval/Retinanet/Eval_Retinanet.py

Modify GradCam_Retinanet.py or visualization.sh.

$ sh visualization.sh

or

$ python ./src/infer/GradCam_Retinanet.py

[email protected]