Project is under active development
Aim of the project is to compare different models to perform CT liver segmentation. Sample output is visible here:
All code is written in PyTorch, wandb for tracking and HuggingFace for transformers. It features few modern solutions:
- Quanet(https://github.com/15029257158/QAU-Net)
- Defednet(https://github.com/SUST-reynole/DefED-Net)
- Vit segmetner(https://github.com/rstrudel/segmenter/tree/master)
- Unet++(https://github.com/MrGiovanni/UNetPlusPlus)
- Polar images(https://github.com/marinbenc/medical-polar-training)
For created experiments ViT segmenter performs best in tumor segmentation and Unet++ for liver segmentation.