Inactive evaluations
method: TencentAILab2017-10-27
Authors: Jingchao Zhou, Hao Wang, Yitong Wang, Zhifeng Li
Description: Our method consists of two networks: text detection network employed in task 1.1 and word recognition network employed in task 1.2. Two networks are integrated with cascade training to achieve superior performance.
method: AlimamaCV2016-05-13
Authors: Quan Chen, Tiezheng Ge, Zhiqiang Zhang, Minghui Li, Kun Gai
Description: We approach this task with a combination of three deep neural networks and a language model. Specifically , a LSTM model is used to accomplish word recognition based on the features generated by a CNN model. The final words are decoded by a bi-gram language model and their locations are refined by a location regression network. Two internal text corpora are involved in the training procedure. For "strongly" and "weakly" version, the given corresponding vocabulary is simply used as the final output filter.
method: Megvii-Image++2016-01-30
Authors: Jia Yu, Xinyu Zhou, Cong Yao, Jianan Wu, Chi Zhang, Shuchang Zhou
Description: A deep neural network based system that consists of two main parts: text detection and word recognition. The detection part is accomplished by a FCN which directly extracts text regions from original images. The recognition part is another neural network that performs whole word recognition.
Date | Method | Recall | Precision | Hmean | |||
---|---|---|---|---|---|---|---|
2017-10-27 | TencentAILab | 90.19% | 90.45% | 90.32% | |||
2016-05-13 | AlimamaCV | 79.49% | 95.81% | 86.89% | |||
2016-01-30 | Megvii-Image++ | 79.21% | 92.53% | 85.35% | |||
2015-10-20 | Deep2Text II+ | 73.92% | 92.27% | 82.08% | |||
2015-04-02 | Stradvision-2 | 73.02% | 83.93% | 78.10% | |||
2015-04-02 | Deep2Text II-1 | 73.37% | 80.97% | 76.98% | |||
2015-04-02 | StradVision-1 | 70.17% | 84.72% | 76.76% | |||
2015-03-30 | Deep2Text I | 61.40% | 83.46% | 70.75% | |||
2015-04-03 | PAL (v1.5) | 61.54% | 65.22% | 63.33% | |||
2015-04-03 | pal_v1.6 | 61.47% | 65.14% | 63.26% | |||
2015-04-02 | pal_v1.0 | 63.14% | 60.65% | 61.87% | |||
2015-04-02 | pal_v1.3 | 62.17% | 60.61% | 61.38% | |||
2015-04-01 | NJU Text (Version3) | 41.31% | 60.12% | 48.97% | |||
2015-04-03 | TextCatcher-2 (LRDE) | 40.26% | 32.11% | 35.73% | |||
2015-04-02 | TextCatcher-1 (LRDE) | 5.01% | 11.58% | 6.99% |