Skip to content
Branch: master
Find file History
andrewluchen and Copybara-Service add test dependencies
PiperOrigin-RevId: 243873292
Latest commit bb8f545 Apr 16, 2019
Permalink
Type Name Latest commit message Commit time
..
Failed to load latest commit information.
data Initial code release for "A General and Adaptive Robust Loss Function… Apr 3, 2019
README.md
adaptive.py
adaptive_test.py
cubic_spline.py Initial code release for "A General and Adaptive Robust Loss Function… Apr 3, 2019
cubic_spline_test.py Initial code release for "A General and Adaptive Robust Loss Function… Apr 3, 2019
distribution.py Initial code release for "A General and Adaptive Robust Loss Function… Apr 3, 2019
distribution_test.py Initial code release for "A General and Adaptive Robust Loss Function… Apr 3, 2019
fit_partition_spline.py Initial code release for "A General and Adaptive Robust Loss Function… Apr 3, 2019
fit_partition_spline_test.py Initial code release for "A General and Adaptive Robust Loss Function… Apr 3, 2019
general.py Initial code release for "A General and Adaptive Robust Loss Function… Apr 3, 2019
general_test.py Initial code release for "A General and Adaptive Robust Loss Function… Apr 3, 2019
requirements.txt
run.sh add test dependencies Apr 16, 2019
util.py Initial code release for "A General and Adaptive Robust Loss Function… Apr 3, 2019
util_test.py
vae.py Initial code release for "A General and Adaptive Robust Loss Function… Apr 3, 2019
wavelet.py
wavelet_test.py Initial code release for "A General and Adaptive Robust Loss Function… Apr 3, 2019

README.md

A General and Adaptive Robust Loss Function

This directory contains reference code for the paper A General and Adaptive Robust Loss Function, Jonathan T. Barron CVPR, 2019

The code is implemented in Tensorflow and the required packages are listed in requirements.txt.

If you'd like this loss, include general.py or adaptive.py and call the loss function. general.py implements the "general" form of the loss, which assumes you are prepared to set and tune hyperparameters yourself, and adaptive.py implements the "adaptive" form of the loss, which tries to adapt the hyperparameters automatically and also includes support for imposing losses in different image representations. The probability distribution underneath the adaptive loss is implemented in distribution.py. Demo code for training different variants of a VAE on Celeb-A as was done in the paper is in vae.py.

You can’t perform that action at this time.