Review and implementation of loss functions for image-to-image neural network
Spring 2020
Master Semester Project
Project: 00384
- Make a complete review of the loss function for image2image applications
- Implement a large collection of loss function as a ImageJ plugin in the framework of deepImage
loss function: SNR, MSE, cross-entropy, SSIM, Jaccard, Wasserstein
.
- Write protocols to test different loss function to train u-net in Jupyter Notebooks
- Application: image denoising, image segmentation
- Supervisors
- Daniel Sage, daniel.sage@epfl.ch, 021 693 51 89, BM 4.135
- Michael Unser, michael.unser@epfl.ch, 021 693 51 75, BM 4.136
- Jaejun Yoo