ASTER: Attentional Scene Text Recognizer with Flexible Rectification
ASTER is an accurate scene text recognizer with flexible rectification mechanism. The research paper can be found here.
The implementation of ASTER reuses code from Tensorflow Object Detection API.
Update
[07/13/2019] A PyTorch port has been made by @ayumiymk.
Correction (10/22/2018)
We have identified a bug we accidentally made in the code that causes only part of SVT images being tested and results in higher results. The bug has been fixed in commit a7e8613. Below are the corrected numbers on SVT. The results are still state-of-the-art, so the conclusions are not affected.
- SVT (50) ASTER: 97.4%; ASTER-A: 96.3%; ASTER-B: 96.1%;
- SVT (None): ASTER: 89.5%; ASTER-A: 80.2%; ASTER-B: 81.6%
Prerequisites
ASTER was developed and tested with TensorFlow r1.4. Higher versions may not work.
ASTER requires Protocol Buffers (version>=2.6). Besides, in Ubuntu 16.04:
sudo apt install cmake libcupti-dev
pip3 install --user protobuf tqdm numpy editdistance
Installation
- Go to
c_ops/and runbuild.shto build the custom operators - Execute
protoc aster/protos/*.proto --python_out=.to build the protobuf files - Add
/path/to/astertoPYTHONPATH, or set this variable for every run
Demo
A demo program is located at aster/demo.py, accompanied with pretrained model files available on our release page. Download model-demo.zip and extract it under aster/experiments/demo/ before running the demo.
To run the demo, simply execute:
python3 aster/demo.py
This will output the recognition result of the demo image and the rectified image.
Training and on-the-fly evaluation
Data preparation scripts for several popular scene text datasets are located under aster/tools. See their source code for usage.
To run the example training, execute
python3 aster/train.py \
--exp_dir experiments/demo \
--num_clones 2
Change the configuration in experiments/aster/trainval.prototxt to configure your own training process.
During the training, you can run a separate program to repeatedly evaluates the produced checkpoints.
python3 aster/eval.py \
--exp_dir experiments/demo
Evaluation configuration is also in trainval.prototxt.
Citation
If you find this project helpful for your research, please cite the following papers:
@article{bshi2018aster,
author = {Baoguang Shi and
Mingkun Yang and
Xinggang Wang and
Pengyuan Lyu and
Cong Yao and
Xiang Bai},
title = {ASTER: An Attentional Scene Text Recognizer with Flexible Rectification},
journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},
volume = {},
number = {},
pages = {1-1},
year = {2018},
}
@inproceedings{ShiWLYB16,
author = {Baoguang Shi and
Xinggang Wang and
Pengyuan Lyu and
Cong Yao and
Xiang Bai},
title = {Robust Scene Text Recognition with Automatic Rectification},
booktitle = {2016 {IEEE} Conference on Computer Vision and Pattern Recognition,
{CVPR} 2016, Las Vegas, NV, USA, June 27-30, 2016},
pages = {4168--4176},
year = {2016}
}
IMPORTANT NOTICE: Although this software is licensed under MIT, our intention is to make it free for academic research purposes. If you are going to use it in a product, we suggest you contact us regarding possible patent issues.
185 Jan 01, 2023
6 Sep 29, 2022
7 Dec 06, 2022
2.4k Jan 08, 2023
20 Dec 15, 2022
128 Dec 04, 2022
3 Jul 13, 2022
198 Dec 29, 2022
3.2k Jan 04, 2023
1.3k Jan 04, 2023
3 Feb 04, 2022
1.5k Dec 28, 2022
140 Dec 23, 2022
1 Nov 08, 2021
138 Dec 26, 2022
26 Jun 17, 2021
31 Nov 22, 2022
2 May 31, 2022
81 Jan 01, 2023
81 Dec 01, 2022