Skip to content
Code for ICCV2019 "Symmetric Cross Entropy for Robust Learning with Noisy Labels"
Python
Branch: master
Clone or download
Latest commit 01c8a89 Nov 13, 2019
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
.idea init Sep 7, 2019
README.md
callback_util.py initial commit Sep 9, 2019
datasets.py initial commit Sep 9, 2019
loss.py added other noisy label learning baselines Sep 9, 2019
models.py initial commit Sep 9, 2019
resnet.py initial commit Sep 9, 2019
train_models.py define P for forward and backward loss Oct 15, 2019
util.py initial commit Sep 9, 2019

README.md

Symmetric Cross Entropy Learning (SL)

Code for ICCV2019 "Symmetric Cross Entropy for Robust Learning with Noisy Labels" https://arxiv.org/abs/1908.06112

Requirements

  • Python 3.5.2
  • Tensorflow 1.10.1
  • Keras 2.2.2

Usage

Simply run the code by python3 train_models.py

It can config with dataset, model, epoch, batchsize, noise_rate, symmetric or asymmetric type noise

The other replication

The reprocuded results by Hanxun Huang are slightly better for all methods. The code can be found here: https://github.com/HanxunHuangLemonBear/SCELoss-Reproduce

You can’t perform that action at this time.