Optical Character Recognition.(2017~)

This repo include code review, paper review in text detection, text recognition, end-to-end.

Paper Review

Text Detection

• Character-level detection
  • Detecting Oriented Text in Natural Images by Linking Segments (B.Shi, 17'CVPR)
  • Robust Wide Baseline Stereo from Maximally Stable Extremal Regions (J.Matas, 17'CVPR)
  • WordSup: Exploiting Word Annotations for Character based Text Detection (H.Hu, 17'ICCV)
  • CRAFT : Character Region Awareness for Text Detection (Y.Baek, 19'CVPR)
• Regression-based detection
• Segmentation-based detection
• Ambiguous
  • EAST : An Efficient and Accurate Scene Text Detector (X.Zhou, 17'CVPR)
  • Deep Direct Regression for Multi-Oriented Scene Text Detection (W.He, 17'ICCV)
  • Single Shot Text Detector with Regional Attention (P.He, 17'ICCV)
  • WeText: Scene Text Detection under Weak Supervision (S.Tian, 17'ICCV)
  • Rotation-sensitive Regression for Oriented Scene Text Detection (M.Liao, 18'CVPR)

Text Reocgnition

• CRNN : An End-to-End Trainable Neural Network for Image-based Sequence Recognition and Its Application to Scene Text Recognition (B.Shi, 17'TPAMI)
• Focusing Attention: Towards Accurate Text Recognition in Natural Images (Z.Cheng, 17'ICCV)
• Deep TextSpotter: An End-to-End Trainable Scene Text Localization and Recognition Framework (M.Busta, 17'ICCV)
• Learning to Read Irregular Text with Attention Mechanisms (X.Yang, 17'IJCAI)
• Edit Probability for Scene Text Recognition (F.Bai, 18'CVPR)
• AON: Towards Arbitrarily-Oriented Text Recognition (Z.Cheng, 18'CVPR)
• ASTER: An Attentional Scene Text Recognizer with Rectification (B.Shi, 18'TPAMI)
• Char-Net: A Character-Aware Neural Network for Distorted Scene Text Recognition (W.Liu, 18'AAAI)
• SqueezedText: A Real-Time Scene Text Recognition by Binary Convolutional Encoder-Decoder Network (Z.Liu, 18'AAAI)
• Show, Attend and Read: A simple and Strong Baseline for Irregular Text Recognition (H.Li, 18'AAAI)

End-to-End

• FOTS: Fast Oriented Text Spotting with a Unified Network (X.Liu, 18'CVPR)

LOSS

• Connectionist Temporal Classification: Labeling Unsegmented Sequence Data with Recurrent Neural Networks(A.Graves, 06'ICML)

Code Implementation

TExt Recognition

2019.03.07 ~ CRNN