Awesome

GuidedMixup

Official PyTorch implementation of "GuidedMixup: An Efficient Mixup Strategy Guided by Saliency Maps" (AAAI'23, Oral) (paper) </br>

Requirements

To install requirements:

pip install -r requirements.txt

Install pairing Algorithm:

python setup.py build_ext --inplace

Training

We provide the code for training the neural network above general classification datasets from PuzzleMix.

Cifar-100

• To reproduce Guided-SR with PreActResNet18 for 1200 epochs, run:

python main.py --dataset cifar100 --data_dir [data_path] --root_dir [save_path] --labels_per_class 500 --arch preactresnet18  --learning_rate 0.1 --momentum 0.9 --decay 0.0001 --epochs 1200 --schedule 400 800 --gammas 0.1 0.1 --train mixup --guided True --condition greedy --mix_prob 0.5 --guided_type sr

• To reproduce Guided-AP with PreActResNet18 for 1200 epochs, run:

python main.py --dataset cifar100 --data_dir [data_path] --root_dir [save_path] --labels_per_class 500 --arch preactresnet18  --learning_rate 0.1 --momentum 0.9 --decay 0.0001 --epochs 1200 --schedule 400 800 --gammas 0.1 0.1 --train mixup --guided True --condition greedy --mix_prob 0.8 --guided_type ap

Tiny-ImageNet

• To reproduce Guided-SR with PreActResNet18 for 1200 epochs, run:

python main.py --dataset tiny-imagenet-200 --data_dir [data_path] --root_dir [save_path] --labels_per_class 500 --arch preactresnet18  --learning_rate 0.2 --momentum 0.9 --decay 0.0001 --epochs 1200 --schedule 600 900 --gammas 0.1 0.1 --train mixup --guided True --condition greedy --mix_prob 0.5 --guided_type sr

• To reproduce Guided-AP with PreActResNet18 for 1200 epochs, run:

python main.py --dataset tiny-imagenet-200 --data_dir [data_path] --root_dir [save_path] --labels_per_class 500 --arch preactresnet18  --learning_rate 0.2 --momentum 0.9 --decay 0.0001 --epochs 1200 --schedule 600 900 --gammas 0.1 0.1 --train mixup --guided True --condition greedy --mix_prob 0.8 --guided_type ap --clean_lam 1.0

Citing this Work and this Implementation

@inproceedings{kang2023guidedmixup,
  title={GuidedMixup: an efficient mixup strategy guided by saliency maps},
  author={Kang, Minsoo and Kim, Suhyun},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  volume={37},
  number={1},
  pages={1096--1104},
  year={2023}
}