- Task 1 - Video mode Localisation
- Task 2 - Video Mode End-to-End
- Task 3 - Still image mode Localisation
- Task 4 - Still image mode Word Recognition
- Task 5 - Still image mode End-to-End
method: NLPR-PAL2017-06-29
Authors: Wenhao He, Fei Yin, Cheng-Lin Liu
Description: this method is based on [1]. We also use images of Chinese and Japanese and Korean from the MLT competition.
[1] Deep Direct Regression for Multi-Oriented Scene Text Detection. arXiv:1703.08289
method: SCUT-DLVClab2017-06-28
Authors: Yuliang Liu, Sheng Zhang, Lianwen Jin
Description: A CNN based method.
method: google_vision_api2017-07-31
Authors: morimoto
Description: google_vision_api(private)
Description Paper Source Code
Date | Method | Recall | Precision | Hmean | |||
---|---|---|---|---|---|---|---|
2017-06-29 | NLPR-PAL | 16.25% | 38.10% | 22.78% | |||
2017-06-28 | SCUT-DLVClab | 11.60% | 46.41% | 18.56% | |||
2017-07-31 | google_vision_api | 4.09% | 14.67% | 6.40% | |||
2017-09-01 | OpenCV + Tesseract | 1.55% | 7.95% | 2.60% |