Robust ReadingCompetition
Challenges

method: Pixel-Anchor 2018-11-16

Authors: Yuan Li, Yuanjie Yu

Description: a novel end-to-end trainable deep neural network framework, named Pixel-Anchor, which combines semantic segmentation and SSD in one network by feature sharing and anchor-level attention mechanism to detect oriented scene text.

method: Amap-CVLab2018-10-29

Authors: Xinran Liu, Hang Zhang, Zhihui Hao, Wei Wang (CASIA)

Description: An instance segmentation based method. This work has been completed by Amap (AutoNavi) Computer Vision Lab, Alibaba Group in cooperation with Institute of Automation, Chinese Academy of Sciences (CASIA).

method: EAST++2017-11-09

Authors: chen zhao cai

Description: this one includes some improvements over EAST(https://github.com/argman/EAST) , this one is a single scale result, but i enlarge the original image by 2X because of so many small texts.

Ranking Table

Description Paper Source Code
DateMethodAverage PrecisionPrecisionRecallHmean
2018-11-16Pixel-Anchor 21.82%40.41%56.69%47.18%
2018-10-29Amap-CVLab30.08%35.48%61.00%44.87%
2017-11-09EAST++27.28%33.57%60.37%43.15%
2018-08-08PKU_VDIG34.97%33.59%58.48%42.67%
2018-01-22FOTS_v224.21%33.62%55.89%41.99%
2018-05-18PSENet_NJU_ImagineLab (single-scale)17.80%31.96%57.29%41.03%
2018-01-22FOTS23.68%32.97%50.24%39.82%
2018-08-23Sogou_MM34.74%29.24%61.32%39.59%
2018-03-12ATL Cangjie OCR31.21%28.76%60.12%38.91%
2017-06-28SCUT_DLVClab125.34%31.48%44.93%37.02%
2018-07-27tongjiu14.06%24.90%56.93%34.65%
2017-06-29SARI_FDU_RRPN_v119.88%22.58%48.02%30.72%
2017-06-28SARI_FDU_RRPN_v024.29%19.91%51.53%28.73%
2018-07-31east-IMPROVE11.47%17.95%42.66%25.27%
2017-06-30TH-DL9.24%16.53%25.97%20.20%
2017-06-30Sensetime OCR27.49%10.93%64.03%18.68%
2017-06-30linkage-ER-Flow6.15%12.13%39.18%18.52%

Ranking Graphic

Ranking Graphic