seg 0.0.1

Image segmentation tools specially for blood vessels.

About

This is a collection of image segmentation projects adjusted for fundus blood vessel segmentation.

1. unet_keras result:

2. unet_torch result:

4. mrsg_torch result (sota):

Run

Provide four flavors:

Module	How to run	Model weights location	Notebooks
model.unet_keras	run_training.py, run_testing.py	test/test_best_weights.h5	1. U-Net - Introduction.ipynb, 2. Fundus Blood Vessel Segmentation.ipynb
model.unet_torch	train.py, test.py	weights/checkpoint.pth	README.md
model.multiple_torch	all codes are inside .ipynb files	best_binclass_model.h5, best_multiclass_model.h5	1. binary segmentation (camvid).ipynb and 2. multiclass segmentation (camvid).ipynb
model.mrsg_torch	python train.py --cfg lib/All.yaml, python inference.py --lib/DRIVE.yaml	results/test/ALL/model/*.pth	README.md

Credits

The following github projects are used:

Module	based on	url
model.unet_keras	Retina blood vessel segmentation with a convolution neural network (U-net)	https://github.com/orobix/retina-unet
model.unet_torch	Retina-Blood-Vessel-Segmentation-in-PyTorch	https://github.com/nikhilroxtomar/Retina-Blood-Vessel-Segmentation-in-PyTorch
model.multiple_torch	Python library with Neural Networks for Image Segmentation based on Keras and TensorFlow.	https://github.com/qubvel/segmentation_models
model.mrsg_torch	Retinal Vessel Segmentation with Pixel-wise Adaptive Filters (ISBI 2022)	https://github.com/Limingxing00/Retinal-Vessel-Segmentation-ISBI2022/

Todo

make a thorough refactor; vessel region detection