Decoupled Attention Network for Text Recognition

By Tianwei Wang, Yuanzhi Zhu, Lianwen Jin, Canjie Luo, Xiaoxue Chen, Yaqiang Wu, Qianying Wang, Mingxiang Cai. AAAI Conference on Artificial Intelligence, 2019

294 citations recorded by the source corpus.

Text recognition has attracted considerable research interests because of its various applications. The cutting-edge text recognition methods are based on attention mechanisms. However, most of attention methods usually suffer from serious alignment problem due to its recurrency alignment operation, where the alignment relies on historical decoding results. To remedy this issue, we propose a decoupled attention network (DAN), which decouples the alignment operation from using historical decoding results. DAN is an effective, flexible and robust end-to-end text recognizer, which consists of three components: 1) a feature encoder that extracts visual features from the input image; 2) a convolutional alignment module that performs the alignment operation based on visual features from the encoder; and 3) a decoupled text decoder that makes final prediction by jointly using the feature map and attention maps. Experimental results show that DAN achieves state-of-the-art performance on multiple text recognition tasks, including offline handwritten text recognition and regular/irregular scene text recognition. Codes will be released.1

Subfields: Computer Vision and Pattern Recognition, Human-Computer Interaction

View Paper: https://www.semanticscholar.org/paper/0708f707ab54888b4b23239a3c3038680b0a2641

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

1. Tao A. Xiang
2. Yi-Zhe Song
3. Ayan Kumar Bhunia
4. Yongxin Yang
5. Timothy M. Hospedales
6. Pinaki Nath Chowdhury
7. Aneeshan Sain
8. Dacheng Tao

In Human-Computer Interaction

1. Tao A. Xiang
2. Wayne Zhang
3. Cha Zhang

Citations in the corpus

• Visual Semantics Allow for Textual Reasoning Better in Scene Text Recognition (2021, unknown): “Some later methods (Wang et al. 2020; Wojna et al. 2017) leverage RNN to strengthen context dependencies between characters.”
• Read Like Humans: Autonomous, Bidirectional and Iterative Language Modeling for Scene Text Recognition (2021, methodology): “The canonical way to perform an LM is auto-regression [44, 3, 45] (Fig.”
• Vectorization and Rasterization: Self-Supervised Learning for Sketch and Handwriting (2021, background): “Handwriting Recognition: Similar to sketches, handwriting recognition has also been heavily explored involving both image space (‘offline’ recognition) [50, 5, 8, 39, 61] and vector space (‘online’ recognition) [26, 20].”
• TrOCR: Transformer-based Optical Character Recognition with Pre-trained Models (2021, methodology): “Most of the methods use sequence models in their encoders after the CNN-based backbone except the FCN encoders in (Wang et al. 2020a), which leads to a significant improvement on CER.”
• RobustScanner: Dynamically Enhancing Positional Clues for Robust Text Recognition (2020, methodology): “Instead of decoupling attention from historical decoding results as done in [49], we propose the position enhancement branch and dynamically fusion module to adaptively adjust the ratio of positional and context clues during decoding.”