Skip to content

A Simple U-net model for Retinal Blood Vessel Segmentation based on tensorflow2

Notifications You must be signed in to change notification settings

DeepTrial/Retina-VesselNet

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

40 Commits
 
 
 
 
 
 

Repository files navigation

[NOTE] Since this project has upgraded to Tensorflow 2.3 on 18th March 2021, you can find old branches which have stopped maintenance from:

VesselNet

A Simple U-net model for Retinal Blood Vessel Segmentation with DRIVE dataset

TestResult

Project Structure

We provide 2 version of projects: jupyter notebook and .py file. The implementation of these two versions is completely consistent. Choose one version and enjoy it!

First to run

For the first time, I recommand to use the version of jupyter notebook, it will give you an intuitive presentation. Different notebooks are made for different purpose:

  • EntireBookForColab.ipynb contains complete part of projects such as process, train, test. Furthermore, it can be run on Google Colab
  • PreprocessIllustartion.ipynb shows some preprocess methods for retinal images.
  • TestAndEvaluation.ipynb is the part for evaluating and testing the model.
  • Training.ipynb is the part for defining and training the model.

Remenber to modify the dataset path according to your setting.

Pretrained Model

  • Dataset can be found here. For Chinese, you can download here.
  • Pretrained model is coming soon...

Train/Test your own image

If you want to test your own image, put your image to the the relevant dir and adjust the patch_size,stride according to your image size.

Citation

This project has been used in:

@inproceedings{2020Eye3DVas,
  title={Eye3DVas: three-dimensional reconstruction of retinal vascular structures by integrating fundus image features},
  author={ Yao Z. and  He K. and Zhou H. and Zhang Z. and Xing C. and Zhou F.},
  booktitle={Frontiers in Optics},
  year={2020},
}

Reference

This project is based on the following 2 papers:

U-Net: Convolutional Networks for Biomedical Image Segmentation

Densely Connected Convolutional Networks