Pytorch implementation of 2D U-Net for image segmentation based on U-Net: Convolutional Networks for Biomedical Image Segmentation .
To run the code, download the images from (https://www.ub.edu/mnms/) and put them in labelled/.
The figures below show some segmentation examples and a boxplot of the dice score. The model achieved a mean (standard deviation) dice score of 0.84 (0.28).
