Joint Optimization Framework for Learning with Noisy Labels
This is an unofficial PyTorch implementation of Joint Optimization Framework for Learning with Noisy Labels. The official Chainer implementation is here.
Requirements
- Python 3.6
- PyTorch 0.4
- torchvision
- progress
- matplotlib
- numpy
Usage
Train the network on the Symmmetric Noise CIFAR-10 dataset (noise rate = 0.7):
First,
python train.py --gpu 0 --out first_sn07 --lr 0.08 --alpha 1.2 --beta 0.8 --percent 0.7
to train and relabel the dataset.
Secondly,
python retrain.py --gpu 0 --out second_sn07 --label first_sn07
to retrain on the relabeled dataset.
Train the network on the Asymmmetric Noise CIFAR-10 dataset (noise rate = 0.4):
First,
python train.py --gpu 0 --out first_an04 --lr 0.03 --alpha 0.8 --beta 0.4 --percent 0.4 --asym
to train and relabel the dataset.
Secondly,
python retrain.py --gpu 0 --out second_an04 --label first_an04
to retrain on the relabeled dataset.
References
- D. Tanaka, D. Ikami, T. Yamasaki and K. Aizawa. "Joint Optimization Framework for Learning with Noisy Labels", in CVPR, 2018.