method: VARCO2019-12-18
Authors: Jusung Lee, Jaemyung Lee, Younghyun Lee, Joonsoo Lee
Affiliation: NCSOFT
Description: This work was supported by Institute of Information & communications Technology Planning & Evaluation (IITP) grant funded by the Korea government(MSIT) (No.1711097855, Text Localization and Recognition for Efficient Digital Contents Analysis)
method: Huawei_GDE_AI2020-09-02
Authors: Wang Jinkun, Liu Chang, Chen Taolve, Zhang Yajun, Xu Dong, Ding Xu,
Affiliation: Huawei GDE Manas
Description: Model: CNN + BiLSTM + CTC
Data: MJSynthText + SynthText (Pretrain); Born-Digital-Images, Focused Scene Text 2013-2015, Incidental Scene Text 2015
method: PhotoOCR2013-04-06
Authors: Alessandro Bissacco, Mark Cummins, Yuval Netzer, Hartmut Neven
Description: Classical over-segmentation and beam search architecture is used. A deep neural network serves as the character classifier, whereas character-ngrams are used for language modelling. Top hypothesis from the beam search are re-ranked using a word-ngram model.
Date | Method | Total Edit distance (case sensitive) | Correctly Recognised Words (case sensitive) | T.E.D. (case insensitive) | C.R.W. (case insensitive) | |||
---|---|---|---|---|---|---|---|---|
2019-12-18 | VARCO | 29.5695 | 95.41% | 14.7965 | 97.15% | |||
2020-09-02 | Huawei_GDE_AI | 42.8155 | 91.80% | 36.7162 | 93.05% | |||
2013-04-06 | PhotoOCR | 103.4071 | 82.21% | 87.1869 | 85.41% | |||
2013-04-05 | MAPS | 186.4451 | 80.40% | 176.6011 | 81.51% | |||
2013-04-05 | PLT | 187.6044 | 80.26% | 178.0789 | 81.38% | |||
2013-04-05 | NESP | 198.1562 | 79.29% | 181.8798 | 80.75% |