Summary of Towards Unified Multi-granularity Text Detection with Interactive Attention, by Xingyu Wan et al.
Towards Unified Multi-granularity Text Detection with Interactive Attention
by Xingyu Wan, Chengquan Zhang, Pengyuan Lyu, Sen Fan, Zihan Ni, Kun Yao, Errui Ding, Jingdong Wang
First submitted to arxiv on: 30 May 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper introduces “Detect Any Text” (DAT), an innovative approach to unify scene text detection, layout analysis, and document page detection into a single end-to-end model. The DAT paradigm efficiently manages text instances at different granularities, including word, line, paragraph, and page levels. A key innovation is the across-granularity interactive attention module, which enhances representation learning by correlating structural information across different text queries. This enables the model to achieve mutually beneficial detection performances across multiple text granularities. Additionally, a prompt-based segmentation module refines detection outcomes for texts of arbitrary curvature and complex layouts. The paper demonstrates state-of-the-art performance on various text-related benchmarks, including multi-oriented/arbitrarily-shaped scene text detection, document layout analysis, and page detection tasks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This research introduces a new way to analyze documents and detect text at different levels (words, lines, paragraphs, and pages). The approach is called “Detect Any Text” (DAT) and it can handle complex layouts and curved text. A key part of DAT is its ability to learn from different types of text and connect them together. This helps the model become better at detecting text in various situations. The paper shows that DAT performs well on various tasks, including recognizing text in documents and analyzing document layouts. |
Keywords
» Artificial intelligence » Attention » Prompt » Representation learning
