/DRIVE

Primary LanguagePython

DRIVE: Digital Retinal Images for Vessel Extraction

Requirements

  • Skimage
  • keras
  • os
  • glob

Preparation

The DRIVE dataset was already download from this challenge page.

All the images already been proccessed and save in file data.

  • afterMask_all: Images which mask out the background.
  • afterMask_blue: The blue channel of images in afterMask_all.
  • afterMask_green: The green channel of images in afterMask_all.
  • afterMask_red: The red channel of images in afterMask_all.
  • GB_image_large: The blue and green channel of images in image_large.
  • image_large: Oringal DRIVE dataset
  • mask: The Ground truth of the vessel segmentation
  • mask_crop: The Ground truth of the vessel segmentation which mask out the background

Training

  • For quickly start, you can simply use the following command to train:

    CUDA_VISIBLE_DEVICES=0,1,2,3 python main.py

UNet Segmentation Result

截圖 2022-11-16 上午11 55 53

截圖 2022-11-16 上午11 54 22

Acknowledgements

This code refers to the following projects:

  1. Retina blood vessel segmentation with a convolution neural network (U-net)