method: HIT2020-05-13
Authors: Sihwan Kim and Taejang Park
Affiliation: Hana Institute of Technology
Description: we present the network architecture to maximize conditional log-likelihood by optimizing the lower bound with a proper approximate posterior that has shown impressive performance in several generative model. In addition, by extending layer of latent variables to multiple layers, the network is able to learn scale robust features with no task specific regularization or data augmentation. We provide a detailed analysis and show the results of three public benchmarks to confirm the efficiency and reliability of the proposed algorithm.
method: Craft++2020-05-19
Authors: Xiangyuan Ren, Anjie Song, Zikun Zhou
Affiliation: Shanghai Jiao Tong University, ShannonAi
Email: xiangyuan_ren@shannonai.com
Description: Out Method is based on CRAFT, with Self Supervised Learning for pretraining and stroke level segmentation for multi-task training
method: CRAFT2018-11-07
Authors: Youngmin Baek, Bado Lee, Dongyoon Han, Sangdoo Yun, and Hwalsuk Lee
Description: We propose a novel text detector called CRAFT. The proposed method effectively detects text area by exploring each character and affinity between characters. To overcome the lack of individual character level annotations, our framework exploits the pseudo character-level bounding boxes acquired by the learned interim model in a weakly-supervised manner.
Clova AI OCR Team, NAVER/LINE Corp.
Date | Method | Recall | Precision | Hmean | |||
---|---|---|---|---|---|---|---|
2020-05-13 | HIT | 95.14% | 98.48% | 96.78% | |||
2020-05-19 | Craft++ | 94.56% | 96.32% | 95.43% | |||
2018-11-07 | CRAFT | 93.06% | 97.43% | 95.20% | |||
2020-12-15 | VARCO | 92.82% | 97.56% | 95.13% | |||
2020-07-31 | TextFuseNet | 91.96% | 97.38% | 94.60% | |||
2020-01-20 | VARCO | 92.47% | 95.02% | 93.73% | |||
2018-04-17 | Ali-Amap-xlab-v4 | 92.13% | 95.18% | 93.63% | |||
2021-01-21 | NCU_MSP | 91.32% | 95.98% | 93.59% | |||
2017-11-15 | Sensetime line-level detection | 89.99% | 97.38% | 93.54% | |||
2020-11-10 | Hancom Vision | 89.81% | 97.26% | 93.39% | |||
2017-11-21 | TencentAILab | 94.37% | 92.33% | 93.34% | |||
2020-01-21 | ChinaUnicom-AI | 91.62% | 94.46% | 93.02% | |||
2019-07-23 | CM-CV&AR | 91.69% | 93.98% | 92.82% | |||
2018-01-22 | FOTS | 90.41% | 95.36% | 92.82% | |||
2018-12-03 | SPCNet_TongJi & UESTC (single scale) | 90.76% | 94.15% | 92.42% | |||
2020-01-06 | NCU_MSP | 91.74% | 93.01% | 92.37% | |||
2017-08-10 | SRC-B-MachineLearningLab-v4 | 90.50% | 94.26% | 92.34% | |||
2017-06-25 | FEN | 89.94% | 94.69% | 92.25% | |||
2017-03-16 | Ali-Amap-xlab-v2 | 91.54% | 92.16% | 91.85% | |||
2017-02-28 | Tencent Youtu | 89.53% | 94.26% | 91.84% | |||
2017-02-17 | NLPR-CASIA | 89.17% | 94.63% | 91.82% | |||
2019-05-29 | crpn.v2 | 89.30% | 94.31% | 91.74% | |||
2019-05-03 | Mask Textspotter | 88.49% | 95.07% | 91.66% | |||
2017-03-23 | RTN | 89.02% | 94.20% | 91.54% | |||
2019-07-12 | stela | 89.37% | 93.80% | 91.53% | |||
2017-12-15 | EPTN-SJTU | 88.95% | 93.55% | 91.19% | |||
2019-02-12 | NCSOFT VISION AI LAB | 91.34% | 90.85% | 91.10% | |||
2016-12-16 | RRPN-4 | 87.31% | 95.19% | 91.08% | |||
2016-12-21 | MSRA_v1 | 88.58% | 93.67% | 91.06% | |||
2016-12-04 | Ali-Amap-xlab | 90.30% | 91.26% | 90.78% | |||
2018-03-05 | HappyCCL | 88.24% | 93.32% | 90.71% | |||
2022-08-26 | HBLAB-OCR | 87.65% | 93.19% | 90.34% | |||
2017-01-22 | SRC-B-MachineLearningLab | 87.07% | 93.28% | 90.07% | |||
2018-05-06 | cvmt_allStepNoLkTree | 89.39% | 90.70% | 90.04% | |||
2016-06-23 | Baidu IDL | 87.11% | 92.83% | 89.88% | |||
2017-02-21 | STDN-2 | 83.89% | 95.15% | 89.17% | |||
2018-05-06 | CVMT_frm62 | 88.00% | 90.34% | 89.16% | |||
2017-02-07 | CAS_HotEye | 84.31% | 94.17% | 88.97% | |||
2017-03-22 | MCLAB_TextBoxes_v2 | 85.57% | 91.87% | 88.61% | |||
2017-02-27 | CTDN-3-PN | 82.72% | 95.18% | 88.51% | |||
2017-07-04 | FPTD | 87.29% | 89.53% | 88.39% | |||
2019-07-17 | AFCTPN | 82.63% | 94.83% | 88.31% | |||
2017-05-02 | TsinghuaOCR | 86.32% | 89.81% | 88.03% | |||
2017-02-25 | CTDN2 | 81.83% | 95.14% | 87.98% | |||
2018-01-04 | crpn | 84.04% | 92.12% | 87.89% | |||
2016-11-08 | CTPN | 82.98% | 92.98% | 87.69% | |||
2021-01-09 | FENET | 79.80% | 96.86% | 87.51% | |||
2019-03-23 | SSP-RPNs with Pytorch | 87.09% | 87.66% | 87.37% | |||
2016-07-18 | SCUT-HCII | 84.22% | 90.66% | 87.32% | |||
2017-07-24 | Jack's TD | 84.93% | 89.49% | 87.15% | |||
2016-03-16 | TextConv+WordGraph | 81.64% | 93.40% | 87.13% | |||
2017-04-06 | ConnLink_pre | 82.37% | 92.40% | 87.10% | |||
2017-03-05 | WeText | 83.07% | 91.06% | 86.88% | |||
2017-04-04 | xmu403 | 83.18% | 90.82% | 86.83% | |||
2018-12-08 | Unicamp-SRBR-v2 | 82.06% | 92.15% | 86.82% | |||
2016-03-16 | TextConv+WordGraph | 81.02% | 93.38% | 86.76% | |||
2017-04-01 | ConnLink | 83.45% | 90.27% | 86.73% | |||
2016-01-27 | SenseTime | 84.22% | 88.84% | 86.47% | |||
2017-05-29 | FCN based network for Text Detection | 84.55% | 87.61% | 86.05% | |||
2016-06-23 | SRC-B-TextProcessingLab | 81.52% | 91.11% | 86.04% | |||
2017-03-01 | STDN | 85.81% | 86.09% | 85.95% | |||
2019-04-19 | CenterText(Single-scale) | 83.14% | 88.90% | 85.93% | |||
2015-11-28 | CASIA_USTB-Cascaded | 82.59% | 89.50% | 85.91% | |||
2016-08-31 | MCLAB_TextBoxes | 83.00% | 89.00% | 85.89% | |||
2015-11-04 | MSER_Binary_CNN | 82.37% | 89.12% | 85.61% | |||
2015-04-03 | StradVision | 80.15% | 90.93% | 85.20% | |||
2016-04-10 | SCUT-HCII | 82.90% | 87.59% | 85.18% | |||
2018-05-06 | cvmt_frm59 | 86.72% | 83.61% | 85.14% | |||
2017-08-15 | MultDet | 83.56% | 86.31% | 84.91% | |||
2017-05-27 | BUPT_NIRC_multi | 79.85% | 90.29% | 84.75% | |||
2017-03-24 | (๑•̀ㅂ•́)و✧ | 85.11% | 84.27% | 84.69% | |||
2017-05-27 | BUPT_NIRC_multi | 79.01% | 90.85% | 84.52% | |||
2017-08-25 | bupt | 80.11% | 88.86% | 84.26% | |||
2015-03-26 | VGGMaxNet_cmb | 77.32% | 92.18% | 84.10% | |||
2015-03-23 | VGGMaxNet_013 | 76.38% | 93.00% | 83.88% | |||
2015-04-02 | VGGMaxNet_025 | 79.76% | 88.42% | 83.87% | |||
2015-10-21 | MCLAB_FCN | 79.65% | 88.40% | 83.80% | |||
2016-01-18 | Text-CNN | 76.29% | 92.69% | 83.69% | |||
2015-03-23 | VGGMaxNet_1.6 | 75.62% | 93.45% | 83.59% | |||
2019-01-31 | ssprpns | 76.29% | 91.81% | 83.33% | |||
2018-12-08 | Unicamp-SRBR-v3 | 76.16% | 91.51% | 83.14% | |||
2017-05-29 | ssd pretrain on synthtext and scut | 79.43% | 86.17% | 82.67% | |||
2014-06-10 | IWRR2014 | 78.65% | 85.89% | 82.11% | |||
2020-03-26 | RRPN R-50 model_final 20200327 | 76.97% | 87.17% | 81.75% | |||
2018-07-17 | 0.956_1280_960 | 75.67% | 88.88% | 81.75% | |||
2014-11-12 | HUST_MCLAB | 76.05% | 87.96% | 81.58% | |||
2017-09-05 | P-SSD.v1 | 79.25% | 83.69% | 81.41% | |||
2019-06-26 | std(single-scale) | 78.05% | 85.02% | 81.38% | |||
2018-07-17 | 123_0.927_1280_960 | 79.11% | 82.96% | 80.99% | |||
2018-07-17 | 0.956_600_1000 | 76.37% | 85.91% | 80.86% | |||
2016-11-13 | RRPN-3 | 72.00% | 90.97% | 80.38% | |||
2015-01-01 | BUCT_YST | 73.88% | 84.64% | 78.90% | |||
2013-08-29 | UMD_IntegratedDisrimination | 69.97% | 89.45% | 78.52% | |||
2019-02-16 | [BKU K15] | 79.03% | 77.96% | 78.49% | |||
2017-04-26 | Cascade Filtering and Grouping | 73.66% | 83.21% | 78.14% | |||
2013-04-07 | USTB_TexStar | 69.28% | 88.80% | 77.83% | |||
2017-06-09 | MPT+Jar | 68.24% | 89.82% | 77.56% | |||
2014-05-13 | SWT | 73.24% | 81.53% | 77.16% | |||
2018-07-03 | ACI | 70.76% | 84.71% | 77.11% | |||
2022-05-08 | 11111 | 77.06% | 76.20% | 76.63% | |||
2013-04-08 | Text_detector_CASIA | 67.27% | 84.97% | 75.09% | |||
2015-07-22 | ZText | 67.05% | 85.01% | 74.97% | |||
2013-12-23 | BayesText | 67.05% | 84.58% | 74.80% | |||
2013-04-05 | TextSpotter | 64.97% | 87.49% | 74.56% | |||
2018-12-08 | Unicamp-SRBR-v1 | 64.40% | 87.40% | 74.16% | |||
2013-04-08 | CASIA_NLPR | 68.82% | 79.26% | 73.67% | |||
2013-04-09 | I2R_NUS_FAR | 70.92% | 75.71% | 73.24% | |||
2017-08-13 | P-SSD.v1 | 66.83% | 79.80% | 72.74% | |||
2017-10-12 | TextFCN V2 | 77.77% | 67.98% | 72.55% | |||
2014-08-18 | DetectText | 68.99% | 75.93% | 72.29% | |||
2015-12-17 | TEST17 | 67.49% | 77.78% | 72.27% | |||
2013-04-08 | I2R_NUS | 69.84% | 73.29% | 71.52% | |||
2018-12-18 | test | 66.76% | 75.76% | 70.98% | |||
2015-03-23 | VGGMaxNet_10 | 55.11% | 97.35% | 70.38% | |||
2013-04-08 | TH-TextLoc | 69.95% | 70.47% | 70.21% | |||
2013-04-06 | Text Detection | 66.05% | 74.50% | 70.02% | |||
2017-04-02 | CNN_Text | 69.17% | 61.99% | 65.38% | |||
2018-12-29 | fast_ret_sh_02 | 54.76% | 77.09% | 64.03% | |||
2017-03-24 | CNN | 68.18% | 59.30% | 63.43% | |||
2015-08-18 | MSER with LocalSWT | 48.11% | 65.93% | 55.63% | |||
2014-09-16 | MSERs | 56.05% | 53.94% | 54.97% | |||
2015-08-18 | MSER | 53.39% | 54.52% | 53.95% | |||
2013-04-25 | Baseline | 35.07% | 60.95% | 44.52% | |||
2013-04-10 | Inkam | 42.54% | 31.73% | 36.35% |