nikhilroxtomar / retina-blood-vessel-segmentation-in-pytorch

This repository contains the code for semantic segmentation of the retina blood vessel on the DRIVE dataset using the PyTorch framework.

Retina-Blood-Vessel-Segmentation-in-PyTorch

This repository contains the code for semantic segmentation of the retina blood vessel on the DRIVE dataset using the PyTorch framework.

The following models are used:

• UNET

Models to be used in future:

• RESUNET
• DEEPLABV3+
• more...

Dataset

The ISIC-2018 dataset is used for this for training the UNET architecture. The dataset contains the 2596 pairs of images and masks. All of these images are of different shapes and contains a variety of skin lesions.

Original Image	Mask Image

Contact

For any query contact below:

• Facebook
• Instagram
• Telegram

retina-blood-vessel-segmentation-in-pytorch's Issues

Bug

RuntimeError: Sizes of tensors must match except in dimension 1. Expected size 56 but got size 64 for tensor number 1 in the list.

You should check the tensors' shape in decoding steps, when you shall crop the tensors' sizes in the encoder block before doing skip connection to tensor at the decoder block.

I am unable to test and train

I am unable to test and train its showing Train: 0 - 0 Test: 0 - 0 0it [00:00, ?it/s] 0it [00:00, ?it/s] in terminal

might be from data path which I provide in addition for that I had download data set in dekstop and taking location directly from there data_path = "/Users/amanbansal/Desktop/dataset"

data_aug file missing

augmentation file is missing from directory