Robust ReadingCompetition
Challenges

method: HoText_v12018-03-05

Authors: Zijian Zhou, Zhizhong Su, Shuangping Huang, Lianwen Jin

Description: An end to end training method, the paper is in preparation.

method: SRC-B-MachineLearningLab2018-03-28

Authors: Xiaobing Wang, Yingying Jiang, Xiangyu Zhu, Yi Yu, Haiyang Guo, Hao Guo, Ning An, Zhenbo Luo

Description: Samsung R&D Institute of China - Beijing, Machine Learning Lab.
Detection is based on R2CNN.
Recognition is CNN-LSTM based method.

method: Baidu VIS2017-10-23

Authors: VIS-OCR

Description: CNN based detection, attention based CNN+RNN recognition.

Ranking Table

Description Paper Source Code
DateMethodRecallPrecisionHmean
2018-03-05HoText_v179.68%91.84%85.33%
2018-03-28SRC-B-MachineLearningLab75.54%95.21%84.24%
2017-10-23Baidu VIS77.80%91.77%84.21%
2018-01-22FOTS75.45%93.61%83.55%
2017-08-02Sensetime V174.19%94.54%83.14%
2017-07-17Baidu IDL v372.85%91.25%81.02%
2017-07-12SRC-B-MachineLearningLab68.32%89.08%77.33%
2017-07-04Baidu IDL v267.36%87.33%76.05%
2016-12-26HUST_MCLAB52.00%97.65%67.86%
2016-07-21Baidu IDL60.81%67.54%64.00%
2016-03-02TextProposals + DictNet37.89%89.84%53.30%
2016-04-05SRC-B-TextProcessingLab40.11%76.42%52.60%
2016-10-20Yunos_Robot1.036.30%67.81%47.29%
2015-11-09Megvii-Image++39.38%57.48%46.74%
2015-04-03Stradvision-232.21%67.92%43.70%
2015-04-02StradVision_v139.77%28.51%33.21%
2015-04-01NJU Text (Version3)24.51%48.80%32.63%
2015-04-02Beam search CUNI +S (decoding for TextSpotter - no postprocessing)15.65%37.83%22.14%
2015-04-02Deep2Text-MO13.82%21.34%16.77%
2015-05-04Baseline OpenCV3.0 + Tesseract8.33%40.90%13.84%
2015-04-02Beam search CUNI (decoding for TextSpotter - spell checked)7.22%81.08%13.26%
2015-04-02Baseline-TextSpotter0.00%0.00%0.00%

Ranking Graphic