Awesome
PyTorch implementation of Res2Net
This repository contains code for Res2Net based on Res2Net: A New Multi-scale Backbone Architecture implemented in PyTorch.
Requirements
- Python 3.6
- PyTorch 1.0
Training
CIFAR-100
- Res2NeXt-29, 6cx24wx4scale:
python train.py --dataset cifar100 --arch res2next29_6cx24wx4scale
- Res2NeXt-29, 8cx25wx4scale:
python train.py --dataset cifar100 --arch res2next29_8cx25wx4scale
- Res2NeXt-29, 6cx24wx6scale:
python train.py --dataset cifar100 --arch res2next29_6cx24wx6scale
- Res2NeXt-29, 6cx24wx4scale-SE:
python train.py --dataset cifar100 --arch res2next29_6cx24wx4scale_se
- Res2NeXt-29, 8cx25wx4scale-SE:
python train.py --dataset cifar100 --arch res2next29_8cx25wx4scale_se
- Res2NeXt-29, 6cx24wx6scale-SE:
python train.py --dataset cifar100 --arch res2next29_6cx24wx6scale_se
ImageNet
- Res2Net-50:
python train.py --dataset imagenet --imagenet-dir </path/to/data> --arch res2net50 --epoch 100 --milestones 30 --weight-decay 1e-4
- Res2Net-101:
python train.py --dataset imagenet --imagenet-dir </path/to/data> --arch res2net101 --epoch 100 --milestones 30 --weight-decay 1e-4
- Res2Net-152:
python train.py --dataset imagenet --imagenet-dir </path/to/data> --arch res2net152 --epoch 100 --milestones 30 --weight-decay 1e-4
- Res2NeXt-50_32x4d:
python train.py --dataset imagenet --imagenet-dir </path/to/data> --arch res2next50_32x4d --epoch 100 --milestones 30 --weight-decay 1e-4
- SE-Res2Net-50:
python train.py --dataset imagenet --imagenet-dir </path/to/data> --arch se_res2net50 --epoch 100 --milestones 30 --weight-decay 1e-4
Results
Coming soon...
<!-- | Model | Error rate | Loss | Error rate (paper) | |:------------------------------------------------|:----------:|:-------:|:------------------:| | WideResNet28-10 baseline | 3.82| 0.158 | 3.89| | WideResNet28-10 +RICAP | **2.82**| 0.141 | **2.85**| | WideResNet28-10 +Random Erasing | 3.18|**0.114**| 4.65| | WideResNet28-10 +Mixup | 3.02| 0.158 | 3.02| Learning curves of loss and accuracy.   -->