LMNL

This is the implementation of our solution (Learning with Real-world Noisy Labels: A Consistency Training Solution) for the first learning and mining with noisy labels challenge in IJCAI-ECAI 2022.

Requirements:

• python=3.8.12
• numpy=1.21.2
• pillow=8.4.0
• pytorch=1.10.2
• requests=2.27.1
• scikit-learn=1.0.1
• scipy=1.7.3
• torchvision=0.11.3

You need to:

1. Download CIFAR-10 and CIFAR-100 datasets into '../data/'.
2. Run the following demos of our solution:

python main.py --dataset cifar10 --noise_type aggre --is_human --lam 0.9 --momentum_1 0.9 --momentum_2 0.9 --momentum_3 0.9 --method 'aggre'
python main.py --dataset cifar10 --noise_type worst --is_human --lam 0.9 --momentum_1 0.9 --momentum_2 0.9 --momentum_3 0.9 --method 'worst'
python main.py --dataset cifar10 --noise_type rand1 --is_human --lam 0.9 --momentum_1 0.9 --momentum_2 0.9 --momentum_3 0.9 --method 'rand1'
python main.py --dataset cifar10 --noise_type rand2 --is_human --lam 0.9 --momentum_1 0.9 --momentum_2 0.9 --momentum_3 0.9 --method 'rand2'
python main.py --dataset cifar10 --noise_type rand3 --is_human --lam 0.9 --momentum_1 0.9 --momentum_2 0.9 --momentum_3 0.9 --method 'rand3'
python main.py --dataset cifar10 --noise_type clean --is_human --lam 0.9 --momentum_1 0.9 --momentum_2 0.9 --momentum_3 0.9 --method 'clean'
python main.py --dataset cifar100 --noise_type noisy100 --is_human --lam 0.9 --momentum_1 0.9 --momentum_2 0.9 --momentum_3 0.9 --method 'noisy100'
python main.py --dataset cifar100 --noise_type clean100 --is_human --lam 0.9 --momentum_1 0.9 --momentum_2 0.9 --momentum_3 0.9 --method 'clean100'

If you have any further questions, please feel free to send an e-mail to: [email protected]. Have fun!