DeepSceneText: A Multi-Scale Transformer Framework for Accurate Text Detection and Recognition
Abstract & Details
Research Area
Computer Science Engineering
Keywords
Text recognition
scene text
DeepSceneText
text detection
multi-scale
transformer
real-life images
blurry
curved text
COCO-Text
ICDAR 2017
SVT
accuracy
speed
real-world applications.
Abstract
We all use text every day — it’s how we share ideas and information. But reading text from real-life pictures (like street signs or labels) is tough because the image might be blurry, dark, or have weird fonts. This project, DeepSceneText, uses smart techniques to make this easier and more accurate.Instead of handling text detection and recognition as two separate jobs, this system does both at the same time. It also looks at the image in multiple sizes (multi-scale), so it can read both big and small text. And it uses transformer technology, which helps find and understand the text, even if it’s curved or not in a straight line.They tested it on standard datasets (like COCO-Text, ICDAR 2017, and SVT), and it worked better than older methods it was more accurate and faster, especially in real-world situations like reading signs or documents.
License
This work is licensed under a Creative
Commons
Attribution-ShareAlike 4.0 International License.
Author Information
| # | Name | Institute / Affiliation |
|---|---|---|
| 1 | Lahor Purva Rajesh | Padmashri Dr Vithalrao Vikhe Patil College of Engineering ,Ahilyanagar |
| 2 | Narang Nishika Manish | Padmashri Dr Vithalrao Vikhe Patil College of Engineering ,Ahilyanagar |
| 3 | Mahajan Bhavika Rajendra | Padmashri Dr Vithalrao Vikhe Patil College of Engineering ,Ahilyanagar |
| 4 | Malve Tanvi Sanjay | Padmashri Dr Vithalrao Vikhe Patil College of Engineering ,Ahilyanagar |
How to Cite
Use the following formats to cite this article in your research.
APA Style
Rajesh, Lahor Purva, Manish, Narang Nishika, Rajendra, Mahajan Bhavika, & Sanjay, Malve Tanvi (2025). DeepSceneText: A Multi-Scale Transformer Framework for Accurate Text Detection and Recognition. International Journal of Advance Research and Innovative Ideas In Education, 11(2), 1548-1554.
MLA Style
Rajesh, Lahor Purva, et al. "DeepSceneText: A Multi-Scale Transformer Framework for Accurate Text Detection and Recognition." International Journal of Advance Research and Innovative Ideas In Education, vol. 11, no. 2, 2025, pp. 1548-1554.
IEEE Style
Lahor Purva Rajesh, Narang Nishika Manish, Mahajan Bhavika Rajendra, and Malve Tanvi Sanjay, "DeepSceneText: A Multi-Scale Transformer Framework for Accurate Text Detection and Recognition," International Journal of Advance Research and Innovative Ideas In Education, vol. 11, no. 2, pp. 1548-1554, 2025.
Vancouver Style
Rajesh Lahor Purva, Manish Narang Nishika, Rajendra Mahajan Bhavika, Sanjay Malve Tanvi. DeepSceneText: A Multi-Scale Transformer Framework for Accurate Text Detection and Recognition. International Journal of Advance Research and Innovative Ideas In Education. 2025;11(2):1548-1554.
Harvard Style
Rajesh, Lahor Purva, Manish, Narang Nishika, Rajendra, Mahajan Bhavika, & Sanjay, Malve Tanvi (2025) 'DeepSceneText: A Multi-Scale Transformer Framework for Accurate Text Detection and Recognition', International Journal of Advance Research and Innovative Ideas In Education, 11(2), pp. 1548-1554.
Chicago Style
Rajesh, Lahor Purva, et al. "DeepSceneText: A Multi-Scale Transformer Framework for Accurate Text Detection and Recognition." International Journal of Advance Research and Innovative Ideas In Education 11, no. 2 (2025): 1548-1554.
Turabian Style
Rajesh, Lahor Purva, et al. "DeepSceneText: A Multi-Scale Transformer Framework for Accurate Text Detection and Recognition." International Journal of Advance Research and Innovative Ideas In Education 11, no. 2 (2025): 1548-1554.
Related Research
EduQuest AI: A Gamified Learning Platform for Interactive and Personalized Digital Education
Download PDF
Distributed Real-Time Stream Analytics and Ingestion Engine for High-Velocity Big Data Architectures
Download PDF