The u-net network I implemented was used to train the dataset for the coal and gangue segmentation of CV course.The best dice value is currently 0.8546 after 100 epochs.
The implementation was mainly referred to https://github.com/milesial/Pytorch-UNet.
If you have any suggestion, please do file issues. I am graceful for any feedback and will do my best to improve this repo.
This repo depends on pytorch>=0.4, available via pip(or conda) install command.
2678701546@qq.com. Please feel free to get in touch with me on any issue related to this repository. For more information about me, please visit my personal website. zhuzhu