ChestBFPN (Work in Progress)

NOTE: We will release the data and finish the code refactoring around the conference time

Bayesian Feature Pyramid Networks for Chest X-ray Segmentation and Automatic Assessment of Cardio-Thoratic Ratio

Requirements

Python 3.*, keras 2.2.4, keras_contrib, segmentation_models, classification_models, PIL, numpy, pandas, cv2

Data structure

1. Download Chest-X-Ray dataset and put it in folder ../input/Chest X-ray-14/img/
2. Download DeepChest archive and put it content in '../input/' folder. It contains:

• folder 'masks_train' - train masks
• folder 'masks_test' - test masks
• file 'train_val_split.pkl' - split on train and validation parts

Usage

python3 train_segmentation_net.py

Citation

If you use it please cite: