• File
  • Strongly Contextualised results
  • Weakly Contextualised results
  • Generic results
  • Evaluation
  • End-to-End recognition
  • Word Spotting

    • method: AIX Lab., LG Electronics 2021-02-05

    Authors: Wonju Lee and Joseph Lim

    Affiliation: AI Lab., Future Tech. Center, CTO, LG Electronics

    Email: wonju2.lee@lge.com

    Description: We reimplemented the end-to-end text spotting method combined the Mask TextSpotter with Deformable ConvNets. We tried to pre-train synthText images for character-level annotations and to fine-tune real world images like following papers for the generalization. We don't use any private data.

    Mask TextSpotter: An End-to-End TrainableNeural Network for Spotting Text with Arbitrary Shapes Minghui Liao, Pengyuan Lyu, Minghang He, Cong Yao, Wenhao Wu, Xiang Bai

    Deformable ConvNets v2: More Deformable, Better Results Xizhou Zhu, Han Hu, Stephen Lin, Jifeng Dai

    Source code

    method: TextBoxes++2017-06-09

    Authors: Minghui Liao, Baoguang Shi, Xiang Bai

    Description: TextBoxes++ and CRNN

    Paper: Liao M, Shi B, Bai X. Textboxes++: A single-shot oriented scene text detector[J]. IEEE Transactions on Image Processing, 2018, 27(8): 3676-3690.

    Source code

    method: FOTS2018-01-22

    Authors: Xuebo Liu, Ding Liang, Shi Yan, Dagui Chen, Yu Qiao, Junjie Yan

    Description: A unified end-to-end trainable Fast Oriented Text Spotting (FOTS) network for simultaneous detection and recognition, sharing computation and visual information among the two complementary tasks.

    FOTS: Fast Oriented Text Spotting with a Unified Network, accepted by CVPR 2018

    Ranking Table

    Description Paper Source Code
    DateMethodRecallPrecisionHmean
    2021-02-05AIX Lab., LG Electronics 92.04%94.20%93.11%
    2017-06-09TextBoxes++89.09%95.44%92.16%
    2018-01-22FOTS87.46%92.93%90.11%

    Ranking Graphic