Focusing Attention: Towards Accurate Text Recognition in Natural Images

By Zhanzhan Cheng, Fan Bai, Yunlu Xu, Gang Zheng, Shiliang Pu, Shuigeng Zhou. IEEE International Conference on Computer Vision, 2017

493 citations recorded by the source corpus.

Scene text recognition has been a hot research topic in computer vision due to its various applications. The state of the art is the attention-based encoder-decoder framework that learns the mapping between input images and output sequences in a purely data-driven way. However, we observe that existing attention-based methods perform poorly on complicated and/or low-quality images. One major reason is that existing methods cannot get accurate alignments between feature areas and targets for such images. We call this phenomenon “attention drift”. To tackle this problem, in this paper we propose the FAN (the abbreviation of Focusing Attention Network) method that employs a focusing attention mechanism to automatically draw back the drifted attention. FAN consists of two major components: an attention network (AN) that is responsible for recognizing character targets as in the existing methods, and a focusing network (FN) that is responsible for adjusting attention by evaluating whether AN pays attention properly on the target areas in the images. Furthermore, different from the existing methods, we adopt a ResNet-based network to enrich deep representations of scene text images. Extensive experiments on various benchmarks, including the IIIT5k, SVT and ICDAR datasets, show that the FAN method substantially outperforms the existing methods.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/45524d7a40435d989579b88b70d25e4d65ac9e3c

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. Xiang Bai
2. Cong Yao
3. Xin He
4. Shangbang Long
5. Pengyuan Lyu
6. Chunhua Shen
7. Minghui Liao
8. Xinggang Wang

Citations in the corpus

• Mask TextSpotter: An End-to-End Trainable Neural Network for Spotting Text with Arbitrary Shapes (2018, unknown): “[8] ResNet, 90k+ST 99.”
• Visual Semantics Allow for Textual Reasoning Better in Scene Text Recognition (2021, unknown): “For example, (Cheng et al. 2017) employ CNN to yield bags of N-grams of text string for output reasoning.”
• Towards Accurate Scene Text Recognition With Semantic Reasoning Networks (2020, methodology): “Most of the recent works have attempted to improve the performance of scene text recognition from the perspective of extracting more robust and effective visual features, such as upgrading the backbone networks [5, 21, 31], adding rectification modules [30, 31, 40, 44] and improving attention mechan”
• Decoupled Attention Network for Text Recognition (2019, methodology): “Traditional attention mechanism often encounters serious alignment problem (Cheng et al. 2017; Bai et al. 2018; Chorowski et al. 2015; Kim, Hori, and Watanabe 2017 (Bahdanau, Cho, and Bengio 2015; Shi et al. 2016) on long text. is because the coupling relationship inevitably leads to error…”
• Scene Text Detection and Recognition: The Deep Learning Era (2018, methodology): “Cheng et al. [13] observed the attention drift problem in existing attention-based methods and proposed an Focus Attention Network (FAN) to attenuate it.”
• ASTER: An Attentional Scene Text Recognizer with Flexible Rectification (2019, methodology): “Our model only falls short on a few results compared to [11] and [26].”
• RobustScanner: Dynamically Enhancing Positional Clues for Robust Text Recognition (2020, methodology): “Later, top-down methods [12,40,15,11,24,41,3,42,57,8,7,25,50,35,1,55,45,16] were proposed, in which text sequences are end-to-end predicted without the single character detection.”
• What Is Wrong With Scene Text Recognition Model Comparisons? Dataset and Model Analysis (2019, methodology): “(V) after the feature extraction stage [23, 24, 4].”
• A Multi-Object Rectified Attention Network for Scene Text Recognition (2019, background): “Cheng et al. [5] used the focusing attention network (FAN) to correct shifts in attentional mechanisms and achieved more accurate position predictions.”
• ESIR: End-To-End Scene Text Recognition via Iterative Image Rectification (2018, methodology): “For the datasets ICDAR2013, [7] achieves the best accuracy but it requires character-level bounding box annotations.”