method: DuXiaoman_OCR2020-05-21
Authors: Hang Yang, Yangchun Wan
Affiliation: Du Xiaoman Financial
Description: Our method is based on Mask RCNN. ResNeXt-152 as our backbone, we first pretrain the model on synthtext 800k, and then finetune on ArT2019,MLT2019 and part of LSVT. Multi-scale training and testing are used to get the final results.
AI-Lab, Du Xiaoman Financial
method: Tencent TEG OCR2019-12-17
Authors: Pei Xu, Hongzhen Wang, Shan Huang, Shen Huang, Qi Ju
Description: This method is based on Mask RCNN. We use resnet152 as backbone and don't use any ensemble methods. We train and test the model in multi scales. We synthesized curved data to pretrain the model. MLT2017 and a small part of LSVT data are used in training.
method: Sogou_OCR2019-11-04
Authors: Liang Wu, Lulu Xu, Long Ma, Xuefeng Su
Description: An arbitrary-shaped text detection method based Mask RCNN, we use resnet-152 as our backbone and IC17MLT as additional training data, multi-scale training and testing are adopted to get the final results.
| Date | Method | Recall | Precision | Hmean | |||
|---|---|---|---|---|---|---|---|
| 2020-05-21 | DuXiaoman_OCR | 79.35% | 87.81% | 83.36% | |||
| 2019-12-17 | Tencent TEG OCR | 81.16% | 85.64% | 83.34% | |||
| 2019-11-04 | Sogou_OCR | 78.49% | 87.94% | 82.95% | |||
| 2019-04-30 | MEGVII_Detection | 76.68% | 89.64% | 82.65% | |||
| 2020-04-22 | Mask R-CNN | 78.55% | 86.43% | 82.30% | |||
| 2019-05-01 | NJU-ImagineLab | 74.21% | 87.35% | 80.24% | |||
| 2019-04-29 | ArtDet-v2 | 73.54% | 86.45% | 79.48% | |||
| 2019-04-26 | baseline_polygon | 75.38% | 82.51% | 78.79% | |||
| 2020-10-01 | TextFuseNet (ResNet-101) | 72.77% | 85.42% | 78.59% | |||
| 2019-04-30 | CUTeOCR | 71.56% | 86.57% | 78.36% | |||
| 2019-04-29 | Sg_ptd | 70.41% | 85.98% | 77.42% | |||
| 2019-04-28 | Alibaba-PAI | 73.25% | 79.18% | 76.10% | |||
| 2019-04-30 | Fudan-Supremind Detection v3 | 71.61% | 79.26% | 75.24% | |||
| 2019-04-29 | SRCB_Art | 70.30% | 80.41% | 75.02% | |||
| 2019-04-30 | A scene text detection method based on maskrcnn | 66.25% | 85.69% | 74.72% | |||
| 2019-04-30 | DMText_art | 66.15% | 85.09% | 74.43% | |||
| 2019-04-30 | TEXT_SNIPER | 71.45% | 76.17% | 73.74% | |||
| 2019-04-28 | CLTDR | 65.92% | 82.58% | 73.32% | |||
| 2019-04-29 | CRAFT | 68.93% | 77.25% | 72.85% | |||
| 2019-04-30 | MaskRCNN_Text | 67.28% | 79.06% | 72.69% | |||
| 2019-04-30 | QAQ | 63.45% | 83.76% | 72.21% | |||
| 2019-04-30 | MaskDet | 67.04% | 76.47% | 71.44% | |||
| 2019-04-24 | fdu_ai | 61.61% | 82.11% | 70.40% | |||
| 2019-04-30 | CCISTD | 60.72% | 81.16% | 69.47% | |||
| 2019-04-30 | Mask RCNN | 73.20% | 65.16% | 68.95% | |||
| 2019-05-01 | TextMask_V1 | 70.58% | 67.33% | 68.92% | |||
| 2019-04-22 | MFTD: Mask Filters for Text Detection | 63.05% | 72.09% | 67.27% | |||
| 2019-04-25 | Art detect by vivo | 57.15% | 80.72% | 66.92% | |||
| 2019-04-29 | PAT-S.Y | 59.64% | 75.72% | 66.72% | |||
| 2019-04-16 | Art_test_baseline_task1 | 62.27% | 71.38% | 66.51% | |||
| 2019-04-30 | DMCA | 64.01% | 69.08% | 66.45% | |||
| 2019-04-30 | TMIS | 53.49% | 86.19% | 66.01% | |||
| 2019-04-22 | mask rcnn | 55.61% | 74.83% | 63.81% | |||
| 2019-05-01 | Unicamp-SRBR-PN-1 | 57.59% | 68.02% | 62.37% | |||
| 2019-04-26 | TP | 51.62% | 78.18% | 62.18% | |||
| 2019-04-28 | Improved Progressive scale expansion Net | 52.24% | 75.88% | 61.88% | |||
| 2019-04-23 | 1 | 59.04% | 57.38% | 58.20% | |||
| 2019-04-27 | TextCohesion_1 | 43.66% | 68.08% | 53.20% | |||
| 2019-04-30 | EM-DATA | 45.11% | 61.34% | 51.99% | |||
| 2019-04-26 | RAST: Robust Arbitrary Shape Text Detector | 35.44% | 71.08% | 47.30% | |||
| 2019-04-30 | MSR | 0.46% | 0.55% | 0.50% |