/Co-Correcting

Primary LanguagePythonApache License 2.0Apache-2.0

Co-Correcting

Official implementation of TMI 2021 paper Co-Correcting: Noise-tolerant Medical Image Classification via collaborative Label Probability Estimation [paper][arxiv]

Requirements:

  • python3.6
  • numpy
  • torch-1.4.0
  • torchvision-0.5.0

Usage

Co-Correcting.py is used for both training a model on dataset with noisy labels and validating it.

Here is an example:

python Co-Correcting.py --dir ./experiment/ --dataset 'mnist' --noise_type sn --noise 0.2 --forget-rate 0.2

or you can train Co-Correcting with .sh:

sh script/mnist.sh