• Home
  • Search
  • How Far Deep Learning Systems for Text Detection and Recognition in Natural Scenes are Affected by Occlusion?
  • Cite Icon7
  • https://doi.org/10.1007/978-3-030-86198-8_15Copy DOI Icon

How Far Deep Learning Systems for Text Detection and Recognition in Natural Scenes are Affected by Occlusion?

  • Jan 1, 2021
  • Aline Geovanna Soares +2 more
Show More
  • Abstract
  • Literature Map
  • References
  • Citations
  • Similar Papers
Abstract

With the rise of deep learning, significant advances in scene text detection and recognition in natural images have been made. However, the severe impact threat to the algorithm’s performance caused by occlusion still consists of an open issue due to the lack of consistent real-world datasets, richer annotations, and evaluations in the specific occlusion problem. Therefore, unlike previous works in this field, our paper addresses occlusions in scene text recognition. The goal is to evaluate the effectiveness and efficiency of existing deep architectures for scene text detection and recognition in various occlusion levels. First, we investigated state-of-the-art scene text and recognition methods and evaluated these current deep architectures performances on ICDAR 2015 dataset without any generated occlusion. Second, we created a methodology to generate large datasets of scene text in natural images with ranges of occlusion between 0 and 100%. From this methodology, we produced the ISTD-OC, a dataset derivated from the ICDAR 2015 database to evaluate deep architectures under different levels of occlusion. The results demonstrated that these existing deep architectures that have achieved state-of-the-art are still far from understanding text instances in a real-world scenario. Unlike the human vision systems, which can comprehend occluded instances by contextual reasoning and association, our extensive experimental evaluations show that current scene text recognition models are inefficient when high occlusions exist in a scene. Nevertheless, for scene text detection, segmentation-based methods, such as PSENet and PAN, are more robust in predicting higher levels of occluded texts. In contrast, methods that detect at the character level, such as CRAFT, are unsatisfactory to heavy occlusions. When it comes to recognition, attention-based methods that benefit contextual information have performed better than CTC-based methods.

Similar Papers
  • Research Article
  • Citations117

Scene text detection and recognition with advances in deep learning: a survey

  • Mar 27, 2019
  • International Journal on Document Analysis and Recognition (IJDAR)
  • Xiyan Liu +2
  • Conference Article
  • Citations12

Occluded Text Detection and Recognition in the Wild

  • May 01, 2022
  • Zobeir Raisi +1
  • Conference Article
  • Citations6

Vietnamese Scene Text Detection and Recognition using Deep Learning: An Empirical Study

  • Jul 29, 2022
  • Nhat Truong Pham +5
  • Conference Article

STV2k

  • Aug 19, 2016
  • Pingping Xiao +3
  • Conference Article
  • Citations68

Dictionary-guided Scene Text Recognition

  • Jun 01, 2021
  • Nguyen Nguyen +6
  • Conference Article
  • Citations6

A Hybrid Deep Neural Network for Urdu Text Recognition in Natural Images

  • Jul 01, 2019
  • Asghar Ali +1
  • Research Article
  • Citations5

A Deep Learning-Based Text Detection and Recognition Approach for Natural Scenes

  • Oct 27, 2022
  • Journal of Circuits, Systems and Computers
  • Xuexiang Li
  • Conference Article
  • Citations4

Research on Text Recognition of Natural Scenes for Complex Situations

  • Jul 22, 2022
  • Wenhua Yu +2
  • Research Article
  • Citations69

SEE: Towards Semi-Supervised End-to-End Scene Text Recognition

  • Apr 27, 2018
  • Proceedings of the AAAI Conference on Artificial Intelligence
  • Christian Bartz +2
  • Conference Article
  • Citations5

Scene text detection and recognition system for visually impaired people in real world

  • Oct 09, 2018
  • Lei Fei +5
  • Conference Article
  • Citations9

STIRER: A Unified Model for Low-Resolution Scene Text Image Recovery and Recognition

  • Oct 26, 2023
  • Minyi Zhao +3
  • Research Article
  • Citations2

A crowdsource based framework for Bengali scene text data collection and detection

  • Nov 15, 2023
  • Computers and Electrical Engineering
  • Md Yearat Hossain +1
  • Book Chapter
  • Citations4

Scene Character and Text Recognition: The State-of-the-Art

  • Jan 01, 2015
  • Chongmu Chen +2
  • Conference Article
  • Citations2

Scene text dataset in Turkish

  • May 01, 2017
  • Nesli Erdogmus
  • Research Article
  • Citations3

Performance Evaluation of Efficient and Accurate Text Detection and Recognition in Natural Scenes Images Using EAST and OCR Fusion

  • Jan 01, 2025
  • International Journal of Advanced Computer Science and Applications
  • Vishnu Kant Soni +5
Cactus Communications logo

Copyright 2026 Cactus Communications. All rights reserved.