/CIFAR-ZOO

PyTorch implementation of CNNs for CIFAR dataset (97.71% on cifar10)

Primary LanguagePythonMIT LicenseMIT

Awesome CIFAR Zoo

This repository contains the pytorch code for multiple CNN architectures and improve methods based on the following papers, hope the implementation and results will helpful for your research!!

Requirements and Usage

Requirements

  • Python >= 3.5
  • PyTorch >= 0.4
  • TensorFlow/Tensorboard (if you want to use the tensorboard for visualization)
  • Other dependencies (pyyaml, easydict, tensorboardX)
pip install -r requirements.txt

Usage

simply run the cmd for the training:

## 1 GPU for lenet
CUDA_VISIBLE_DEVICES=0 python -u train.py --work-path ./experiments/cifar10/lenet

## resume from ckpt
CUDA_VISIBLE_DEVICES=0 python -u train.py --work-path ./experiments/cifar10/lenet --resume

## 2 GPUs for resnet1202
CUDA_VISIBLE_DEVICES=0,1 python -u train.py --work-path ./experiments/cifar10/preresnet1202

## 4 GPUs for densenet190bc
CUDA_VISIBLE_DEVICES=0,1,2,3 python -u train.py --work-path ./experiments/cifar10/densenet190bc

We use yaml file config.yaml to save the parameters, check any files in ./experimets for more details.
You can see the training curve via tensorboard, tensorboard --logdir path-to-event --port your-port.
The training log will be dumped via logging, check log.txt in your work path.

Results on CIFAR

Vanilla architectures

architecture params batch size epoch C10 test acc (%) C100 test acc (%)
Lecun 62K 128 250 67.46 34.10
alexnet 2.4M 128 250 75.56 38.67
vgg19 20M 128 250 93.00 72.07
preresnet20 0.27M 128 250 91.88 67.03
preresnet110 1.7M 128 250 94.24 72.96
preresnet1202 19.4M 128 250 94.74 75.28
densenet100bc 0.76M 64 300 95.08 77.55
densenet190bc 25.6M 64 300 96.11 82.59
resnext29_16x64d 68.1M 128 300 95.94 83.18
se_resnext29_16x64d 68.6M 128 300 96.15 83.65
cbam_resnext29_16x64d 68.7M 128 300 96.27 83.62
ge_resnext29_16x64d 70.0M 128 300 96.21 83.57

With additional regularization

PS: the default data augmentation methods are RandomCrop + RandomHorizontalFlip + Normalize,
and the means which additional method be used. 🍰

architecture epoch cutout mixup C10 test acc (%)
preresnet20 250 91.88
preresnet20 250 92.57
preresnet20 250 92.71
preresnet20 250 92.66
preresnet110 250 94.24
preresnet110 250 94.67
preresnet110 250 94.94
preresnet110 250 95.66
se_resnext29_16x64d 300 96.15
se_resnext29_16x64d 300 96.60
se_resnext29_16x64d 300 96.86
se_resnext29_16x64d 300 97.03
cbam_resnext29_16x64d 300 97.16
ge_resnext29_16x64d 300 97.19
-- -- -- -- --
shake_resnet26_2x64d 1800 96.94
shake_resnet26_2x64d 1800 97.20
shake_resnet26_2x64d 1800 97.42
shake_resnet26_2x64d 1800 97.71

PS: shake_resnet26_2x64d achieved 97.71% test accuracy with cutout and mixup!!
It's cool, right?

With different LR scheduler

architecture epoch step decay cosine htd(-6,3) cutout mixup C10 test acc (%)
preresnet20 250 91.88
preresnet20 250 92.13
preresnet20 250 92.44
preresnet20 250 93.30
preresnet110 250 94.24
preresnet110 250 94.48
preresnet110 250 94.82
preresnet110 250 95.88

Acknowledgments

Provided codes were adapted from

Feel free to contact me if you have any suggestions or questions, issues are welcome,
create a PR if you find any bugs or you want to contribute. 😊