Awesome

SwinTextSpotter

This is the pytorch implementation of Paper: SwinTextSpotter: Scene Text Spotting via Better Synergy between Text Detection and Text Recognition (CVPR 2022). The paper is available at this link.

We use the models pre-trained on ImageNet. The ImageNet pre-trained SwinTransformer backbone is obtained from SwinT_detectron2.

News

2024.04.09 We release a new text spotting pipeline Bridge Text Spotting that combines the advantages of end-to-end and two-step text spotting. Code

2023.08.22 We release a strong text spotting model ESTextSpotter that achieves explicit synergy on text spotting tasks. Code

Models

SWINTS-swin-english-pretrain [config] | model_Google Drive | model_BaiduYun PW: 954t

SWINTS-swin-Total-Text [config] | model_Google Drive | model_BaiduYun PW: tf0i

SWINTS-swin-ctw [config] | model_Google Drive | model_BaiduYun PW: 4etq

SWINTS-swin-icdar2015 [config] | model_Google Drive | model_BaiduYun PW: 3n82

SWINTS-swin-ReCTS [config] | model_Google Drive | model_BaiduYun PW: a4be

SWINTS-swin-vintext [config] | model_Google Drive | model_BaiduYun PW: slmp

Installation

Python=3.8
PyTorch=1.8.0, torchvision=0.9.0, cudatoolkit=11.1
OpenCV for visualization

Steps

Install the repository (we recommend to use Anaconda for installation.)

conda create -n SWINTS python=3.8 -y
conda activate SWINTS
conda install pytorch==1.8.0 torchvision==0.9.0 torchaudio==0.8.0 cudatoolkit=11.1 -c pytorch -c conda-forge
pip install opencv-python
pip install scipy
pip install shapely
pip install rapidfuzz
pip install timm
pip install Polygon3
git clone https://github.com/mxin262/SwinTextSpotter.git
cd SwinTextSpotter
python setup.py build develop

dataset path

datasets
|_ totaltext
|  |_ train_images
|  |_ test_images
|  |_ totaltext_train.json
|  |_ weak_voc_new.txt
|  |_ weak_voc_pair_list.txt
|_ mlt2017
|  |_ train_images
|  |_ annotations/icdar_2017_mlt.json
.......

Downloaded images

ICDAR2017-MLT [image]
Syntext-150k:
- Part1: 94,723 [dataset]
- Part2: 54,327 [dataset]
ICDAR2015 [image]
ICDAR2013 [image]
Total-Text_train_images [image]
Total-Text_test_images [image]
ReCTs [images&label] PW: 2b4q
LSVT [images&label] PW: 9uh1
ArT [images&label] PW: 2865
SynChinese130k [images][label]
Vintext_images [image]

Downloaded label[Google Drive] [BaiduYun] PW: 46vd

Downloader lexicion[Google Drive] and place it to corresponding dataset.

You can also prepare your custom dataset following the example scripts. [example scripts]

Totaltext

To evaluate on Total Text, CTW1500, ICDAR2015, first download the zipped annotations and unzip it

Pretrain SWINTS (e.g., with Swin-Transformer backbone)

python projects/SWINTS/train_net.py \
  --num-gpus 8 \
  --config-file projects/SWINTS/configs/SWINTS-swin-pretrain.yaml

Fine-tune model on the mixed real dataset

python projects/SWINTS/train_net.py \
  --num-gpus 8 \
  --config-file projects/SWINTS/configs/SWINTS-swin-mixtrain.yaml

Fine-tune model

python projects/SWINTS/train_net.py \
  --num-gpus 8 \
  --config-file projects/SWINTS/configs/SWINTS-swin-finetune-totaltext.yaml

Evaluate SWINTS (e.g., with Swin-Transformer backbone)

python projects/SWINTS/train_net.py \
  --config-file projects/SWINTS/configs/SWINTS-swin-finetune-totaltext.yaml \
  --eval-only MODEL.WEIGHTS ./output/model_final.pth

Visualize the detection and recognition results (e.g., with ResNet50 backbone)

python demo/demo.py \
  --config-file projects/SWINTS/configs/SWINTS-swin-finetune-totaltext.yaml \
  --input input1.jpg \
  --output ./output \
  --confidence-threshold 0.4 \
  --opts MODEL.WEIGHTS ./output/model_final.pth

Example results:

Acknowlegement

Adelaidet, Detectron2, ISTR, SwinT_detectron2, Focal-Transformer and MaskTextSpotterV3.

Citation

If our paper helps your research, please cite it in your publications:

@article{huang2022swints,
  title = {SwinTextSpotter: Scene Text Spotting via Better Synergy between Text Detection and Text Recognition},
  author = {Mingxin Huang and YuLiang liu and Zhenghao Peng and Chongyu Liu and Dahua Lin and Shenggao Zhu and Nicholas Yuan and Kai Ding and Lianwen Jin},
  journal={arXiv preprint arXiv:2203.10209},
  year = {2022}
}

Copyright

For commercial purpose usage, please contact Dr. Lianwen Jin: eelwjin@scut.edu.cn