method: Tencent TEG OCR2020-03-15
Authors: Pei Xu, Shan Huang, Hongzhen Wang, Shen Huang, Qi Ju.
Description: We reimplemented the standalone recognition method according to the end-to-end text spotting code released by the Mask TextSpotter[TPAMI]. It is a seq-to-seq method based on 2D attention. We synthesize curved text images for pretraining by the method of VGG synthtext. We add public dataset including icdar2013-2015, CUTE, SVT, IIIT5k, RCTW2017, LSVT to finetune and don't use any private data.
method: AntMaskText32023-05-06
Authors: JIe Wang, Tao Huang, Taoxu
Affiliation: Ant-Group
Description: Based on Maskrcnn + SPN + SAM, we select on MaskTextSpotterv3 as our baseline. we construct a dictionary with 5800 classes. Some public datasets are used in our experiments, including icdar2013, icdar2015, LSVT, RCTW, part of MLT, and ReCTS. In the 2D-attention module, we improve the representation ability to recognize the chinese characters.
Mask TextSpotter v3: Segmentation Proposal Network for Robust Scene Text Spotting
method: baseline_0.5_class_54352019-04-28
Authors: Jinjin Zhang, Beihang University
Description: instance segment based method for text detection and attention based method for text recognition with threshold 0.5 and 5435 classes. Data augmentation and extra datasets including LSVT, ICDAR2017, COCO-Text, RECTS are used for text recognition.
Date | Method | Recall | Precision | Hmean | 1-NED | |||
---|---|---|---|---|---|---|---|---|
2020-03-15 | Tencent TEG OCR | 60.33% | 63.66% | 61.95% | 64.13% | |||
2023-05-06 | AntMaskText3 | 52.04% | 64.69% | 57.68% | 59.47% | |||
2019-04-28 | baseline_0.5_class_5435 | 47.98% | 52.56% | 50.17% | 54.91% |