Paper

Mumford–Shah Loss Functional for Image Segmentation With Deep Learning
- Authors: Boah Kim and Jong Chul Ye
- published in IEEE Transactions on Image Processing (TIP)

Implementation

A PyTorch implementation of deep-learning-based segmentation based on original cycleGAN code. [https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix] (*Thanks for Jun-Yan Zhu and Taesung Park, and Tongzhou Wang.)

Training: LiTS_train_unet.py which is handled by scripts/LiTS_train_unet.sh
A code for Mumford-Shah loss functional is in models/loss.py.
- 'levelsetLoss' and 'gradientLoss2d' classes compose our Mumford-Shah loss function.

Name		Name	Last commit message	Last commit date
Latest commit History 3 Commits
models		models
options		options
scripts		scripts
util		util
LiTS_getDatabase_unet.py		LiTS_getDatabase_unet.py
LiTS_train_unet.py		LiTS_train_unet.py
README.md		README.md
image_util_unet.py		image_util_unet.py