ieee820 / keras-BTS-DSN

This is a Keras (2.2.4) implementation of BTS-DSN

keras-BTS-DSN

Introduction

This is a Keras (2.2.4) implementation of BTS-DSN, which is a Neural Network for Retinal Vessel Segmentation. The original code is written by Caffe.

Environment

keras = 2.2.4

tensorflow = 1.12.0

python = 3.6

Training BTS-DSN

1. Download VGG16 pretrained model.

2. Prepare the training and testing set.

   python prepare_dataset_hdf5_for_2D.py  # the source code for generating training dataset data in hdf5 format
   python prepare_dataset_hdf5_for_2D_test.py # the source code for generating testing dataset data in hdf5 format

3. Run the training script.

   python deep_supervised.py

Testing BTS-DSN

Run the testing script.

python 2D_new_run_test.py

Acknowledgement

BTS-DSN

SVS-net

License