Scene Text Detection using Hyperbolic Tangent Binarization and CLIP

Deep learning-based scene text detection has advanced significantly and has promising application prospects. The binarization of hyperbolic tangent network plus (HTBNet++) is designed to solve the problem of missing small text and dense text in scene images. By designing the backbone network based on CLIP, pretrained text and image features were introduced to achieve better feature extraction. At the same time, text and image prompts play a key role in feature fusion so that the features of text and image pairs can be better applied to text detection. In addition, an auxiliary segmentation loss was designed to guide the network to better perform reverse gradient propagation during the training process. Experimental results on Total-Text and TD500 datasets demonstrate that the proposed method significantly enhances text detection accuracy and robustness.

• Publication year

  2024

• Venue

  International Conference on Robotics, Intelligent Control and Artificial Intelligence

• Publication date

  2024-12-06

• Fields of study

  Not labeled

• Identifiers
  DOI 10.1109/RICAI64321.2024.10910961
• External record

  Open on Semantic Scholar

• Source metadata

  Semantic Scholar

• No claims are published for this paper.

• No concepts are published for this paper.

• (HTBNet)Arbitrary Shape Scene Text Detection with Binarization of Hyperbolic Tangent and Cross-Entropy

  2024cited by this paper
• Scene Chinese Recognition with Local and Global Attention

  2024cited by this paper
• DenseTextPVT: Pyramid Vision Transformer with Deep Multi-Scale Feature Refinement Network for Dense Text Detection

  2023cited by this paper
• Background-Insensitive Scene Text Recognition with Text Semantic Segmentation

  2022cited by this paper
• Real-Time Scene Text Detection With Differentiable Binarization and Adaptive Scale Fusion

  2022cited by this paper
• Quo vadis artificial intelligence?

  2022cited by this paper
• A review of natural scene text detection methods

  2021cited by this paper
• Learning Transferable Visual Models From Natural Language Supervision

  2021cited by this paper
• Street View Text Recognition With Deep Learning for Urban Scene Understanding in Intelligent Transportation Systems

  2020cited by this paper
• Deep Relational Reasoning Graph Network for Arbitrary Shape Text Detection

  2020cited by this paper
• SegLink++: Detecting Dense and Arbitrary-shaped Scene Text by Instance-aware Component Grouping

  2019cited by this paper
• MSR: Multi-Scale Shape Regression for Scene Text Detection

  2019cited by this paper
• Arbitrary Shape Scene Text Detection With Adaptive Text Region Representation

  2019cited by this paper
• Learning Shape-Aware Embedding for Scene Text Detection

  2019cited by this paper
• Efficient and Accurate Arbitrary-Shaped Text Detection With Pixel Aggregation Network

  2019cited by this paper
• Real-time Scene Text Detection with Differentiable Binarization

  2019cited by this paper
• PixelLink: Detecting Scene Text via Instance Segmentation

  2018cited by this paper
• TextSnake: A Flexible Representation for Detecting Text of Arbitrary Shapes

  2018cited by this paper
• Shape Robust Text Detection With Progressive Scale Expansion Network

  2018cited by this paper
• TextBoxes++: A Single-Shot Oriented Scene Text Detector

  2018cited by this paper
• Detecting Oriented Text in Natural Images by Linking Segments

  2017cited by this paper
• Total-Text: A Comprehensive Dataset for Scene Text Detection and Recognition

  2017cited by this paper
• EAST: An Efficient and Accurate Scene Text Detector

  2017cited by this paper
• An End-to-End Trainable Neural Network for Image-Based Sequence Recognition and Its Application to Scene Text Recognition

  2015cited by this paper
• Detecting texts of arbitrary orientations in natural images

  2012cited by this paper

• No citing papers are available for this paper.