CSwin

CSWin Transformer: A General Vision Transformer Backbone with Cross-Shaped Windows, arxiv

PaddlePaddle training/validation code and pretrained models for CSWin Transformer.

The official pytorch implementation is here.

This implementation is developed by PaddleViT.

CSWin Transformer Model Overview

Update

• Update (2021-09-27): Model FLOPs and # params are uploaded.
• Update (2021-08-11): Code is released and ported weights are uploaded.

Models Zoo

Model	Acc@1	Acc@5	#Params	FLOPs	Image Size	Crop_pct	Interpolation	Link
cswin_tiny_224	82.81	96.30	22.3M	4.2G	224	0.9	bicubic	google/baidu(4q3h)
cswin_small_224	83.60	96.58	34.6M	6.5G	224	0.9	bicubic	google/baidu(gt1a)
cswin_base_224	84.23	96.91	77.4M	14.6G	224	0.9	bicubic	google/baidu(wj8p)
cswin_base_384	85.51	97.48	77.4M	43.1G	384	1.0	bicubic	google/baidu(rkf5)
cswin_large_224	86.52	97.99	173.3M	32.5G	224	0.9	bicubic	google/baidu(b5fs)
cswin_large_384	87.49	98.35	173.3M	96.1G	384	1.0	bicubic	google/baidu(6235)

*The results are evaluated on ImageNet2012 validation set.

Notebooks

We provide a few notebooks in aistudio to help you get started:

*(coming soon)*

Requirements

• Python>=3.6
• yaml>=0.2.5
• PaddlePaddle>=2.1.0
• yacs>=0.1.8

Data

ImageNet2012 dataset is used in the following folder structure:

│imagenet/
├──train/
│  ├── n01440764
│  │   ├── n01440764_10026.JPEG
│  │   ├── n01440764_10027.JPEG
│  │   ├── ......
│  ├── ......
├──val/
│  ├── n01440764
│  │   ├── ILSVRC2012_val_00000293.JPEG
│  │   ├── ILSVRC2012_val_00002138.JPEG
│  │   ├── ......
│  ├── ......

Usage

To use the model with pretrained weights, download the .pdparam weight file and change related file paths in the following python scripts. The model config files are located in ./configs/.

For example, assume the downloaded weight file is stored in ./cswin_base_224.pdparams, to use the cswin_base_224 model in python:

from config import get_config
from cswin import build_cswin as build_model
# config files in ./configs/
config = get_config('./configs/cswin_base_224.yaml')
# build model
model = build_model(config)
# load pretrained weights
model_state_dict = paddle.load('./cswin_base_224.pdparams')
model.set_dict(model_state_dict)

Evaluation

To evaluate CSWin model performance on ImageNet2012 with a single GPU, run the following script using command line:

sh run_eval.sh

or

CUDA_VISIBLE_DEVICES=0 \
python main_single_gpu.py \
    -cfg=./configs/cswin_base_224.yaml \
    -dataset=imagenet2012 \
    -batch_size=16 \
    -data_path=/path/to/dataset/imagenet/val \
    -eval \
    -pretrained=/path/to/pretrained/model/cswin_base_224  # .pdparams is NOT needed

Run evaluation using multi-GPUs:

sh run_eval_multi.sh

or

CUDA_VISIBLE_DEVICES=0,1,2,3 \
python main_multi_gpu.py \
    -cfg=./configs/cswin_base_224.yaml \
    -dataset=imagenet2012 \
    -batch_size=16 \
    -data_path=/path/to/dataset/imagenet/val \
    -eval \
    -pretrained=/path/to/pretrained/model/cswin_base_224  # .pdparams is NOT needed

Training

To train the CSWin model on ImageNet2012 with single GPU, run the following script using command line:

sh run_train.sh

or

CUDA_VISIBLE_DEVICES=0 \
python main_single_gpu.py \
  -cfg=./configs/cswin_base_224.yaml \
  -dataset=imagenet2012 \
  -batch_size=32 \
  -data_path=/path/to/dataset/imagenet/train \

Run training using multi-GPUs:

sh run_train_multi.sh

or

CUDA_VISIBLE_DEVICES=0,1,2,3 \
python main_multi_gpu.py \
    -cfg=./configs/cswin_base_224.yaml \
    -dataset=imagenet2012 \
    -batch_size=16 \
    -data_path=/path/to/dataset/imagenet/train \

Visualization Attention Map

(coming soon)

Reference

@article{dong2021cswin,
  title={CSWin Transformer: A General Vision Transformer Backbone with Cross-Shaped Windows},
  author={Dong, Xiaoyi and Bao, Jianmin and Chen, Dongdong and Zhang, Weiming and Yu, Nenghai and Yuan, Lu and Chen, Dong and Guo, Baining},
  journal={arXiv preprint arXiv:2107.00652},
  year={2021}
}