- Task 1 - Text Localization
- Task 2 - Script identification
- Task 3 - Joint text detection and script identification
method: AntAI-Cognition2020-04-22
Authors: Qingpei Guo, Yudong Liu, Pengcheng Yang, Yonggang Li, Yongtao Wang, Jingdong Chen, Wei Chu
Affiliation: Ant Group & PKU
Email: qingpei.gqp@antgroup.com
Description: We are from Ant Group & PKU. Our approach is an ensemble method with three text detection models. The text detection models mainly follow the MaskRCNN framework[1], with different backbones(ResNext101-64x4d[2], CBNet[3], ResNext101-32x32d_wsl[4]) used. GBDT[5] is trained to normalize confidence scores and select quadrilateral boxes with the highest quality from all text detection models' outputs. Multi-scale training and testing are adopted for all basic models. For the training set, we also add ICDAR19 MLT datasets, both training & validation sets are used to get the final result.
[1] He K, Gkioxari G, Dollár P, et al. Mask r-cnn[C]//Proceedings of the IEEE international conference on computer vision. 2017: 2961-2969. [2] Xie S, Girshick R, Dollár P, et al. Aggregated residual transformations for deep neural networks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 1492-1500. [3] Liu Y, Wang Y, Wang S, et al. Cbnet: A novel composite backbone network architecture for object detection[J]. arXiv preprint arXiv:1909.03625, 2019. [4] Mahajan D, Girshick R, Ramanathan V, et al. Exploring the limits of weakly supervised pretraining[C]//Proceedings of the European Conference on Computer Vision (ECCV). 2018: 181-196. [5] Ke G, Meng Q, Finley T, et al. Lightgbm: A highly efficient gradient boosting decision tree[C]//Advances in neural information processing systems. 2017: 3146-3154.
method: TH2020-04-16
Authors: Tsinghua University and Hyundai Motor Group AIRS Company
Email: Shanyu Xiao: xiaosy19@mails.tsinghua.edu.cn
Description: We have built an end-to-end scene text spotter based on Mask R-CNN & Transformer. The ResNeXt-101 backbone and multiscale training/testing are used.
method: Sogou_OCR2019-11-08
Authors: Xudong Rao, Lulu Xu, Long Ma, Xuefeng Su
Description: An arbitrary-shaped text detection method based on Mask R-CNN, we use resnext-152 as our backbone, multi-scale training and testing are adopted to get the final results.
Date | Method | Hmean | Precision | Recall | Average Precision | |||
---|---|---|---|---|---|---|---|---|
2020-04-22 | AntAI-Cognition | 84.36% | 85.92% | 82.86% | 78.41% | |||
2020-04-16 | TH | 84.19% | 87.21% | 81.38% | 78.36% | |||
2019-11-08 | Sogou_OCR | 83.74% | 87.22% | 80.54% | 76.87% | |||
2019-08-08 | JDAI | 82.50% | 84.24% | 80.82% | 78.13% | |||
2019-06-02 | NJU-ImagineLab | 82.40% | 83.20% | 81.62% | 77.98% | |||
2019-05-30 | PMTD | 81.88% | 84.15% | 79.74% | 76.74% | |||
2019-06-11 | 4Paradigm-Data-Intelligence | 81.07% | 81.85% | 80.30% | 65.57% | |||
2019-05-08 | Baidu-VIS | 80.75% | 83.95% | 77.79% | 65.11% | |||
2019-05-23 | 4Paradigm-Data-Intelligence | 80.62% | 81.80% | 79.47% | 64.81% | |||
2019-03-23 | PMTD | 80.49% | 83.04% | 78.09% | 74.19% | |||
2019-12-13 | BDN | 78.69% | 79.18% | 78.20% | 61.94% | |||
2018-10-29 | Amap-CVLab | 77.20% | 79.64% | 74.91% | 70.51% | |||
2019-03-29 | GNNets (single scale) | 76.90% | 82.75% | 71.83% | 64.55% | |||
2018-11-15 | USTC-NELSLIP | 76.88% | 77.47% | 76.30% | 71.29% | |||
2018-11-28 | CRAFT | 76.71% | 81.30% | 72.60% | 59.13% | |||
2018-11-20 | Pixel-Anchor | 76.04% | 83.58% | 69.75% | 58.14% | |||
2018-05-18 | PSENet_NJU_ImagineLab (single-scale) | 74.94% | 78.55% | 71.65% | 56.39% | |||
2018-12-04 | SPCNet_TongJi & UESTC (multi scale) | 74.29% | 77.17% | 71.63% | 55.07% | |||
2017-11-09 | EAST++ | 73.88% | 78.90% | 69.45% | 56.21% | |||
2019-07-15 | stela | 73.72% | 78.67% | 69.35% | 64.14% | |||
2019-01-08 | ALGCD_CP | 73.18% | 76.52% | 70.12% | 56.41% | |||
2018-03-12 | ATL Cangjie OCR | 73.04% | 75.47% | 70.76% | 65.19% | |||
2018-12-03 | SPCNet_TongJi & UESTC (single scale) | 68.08% | 68.13% | 68.02% | 46.20% | |||
2018-12-05 | EPTN-SJTU | 67.69% | 73.30% | 62.87% | 49.91% | |||
2019-05-30 | Thesis-SE | 66.83% | 72.60% | 61.92% | 47.65% | |||
2017-06-28 | SCUT_DLVClab1 | 63.21% | 76.76% | 53.73% | 48.24% | |||
2017-06-29 | SARI_FDU_RRPN_v1 | 62.25% | 68.90% | 56.77% | 51.75% | |||
2017-06-28 | SARI_FDU_RRPN_v0 | 59.65% | 65.01% | 55.12% | 48.79% | |||
2017-06-30 | Sensetime OCR | 57.74% | 48.74% | 70.83% | 60.84% | |||
2017-06-30 | TH-DL | 43.38% | 62.62% | 33.18% | 29.50% | |||
2017-06-30 | linkage-ER-Flow | 29.78% | 36.84% | 24.99% | 13.62% |