Scene text detection improves by learning from distant blurred text
Back-Tracking from Clarity: Self-Learning to See Text from Afar
Computer Vision and Pattern Recognition
Summary
Detecting text in photos or videos from far away is hard because the text looks tiny and blurry. The authors introduced a new method that teaches a computer to spot this difficult text by using clear and easy-to-read text seen later in videos as a guide. They track the text backward through the video to label earlier blurry frames automatically, allowing the computer to learn without needing people to label data. They also created a new multilingual outdoor dataset to help train and test these systems. Their method helps computers recognize text earlier and more accurately in various scenes and languages.
What this means in practice
- •For mobile app developers: Improve mobile OCR apps to recognize small or blurry distant text in videos without extra manual training data.
- •For autonomous vehicle teams: Enhance text detection in driver assistance systems to read road signs or text from a distance under difficult conditions.
Authors
Duc-Tri Tran, Phi Le Nguyen, Minh Hoai
Abstract
We propose a self-supervised framework designed to enhance the capability of scene text detectors in identifying and recognizing text in scenarios where instances are shown at significant distances, typically small, blurred, and frequently missed by conventional models. Our approach leverages the high-fidelity performance of existing text spotting models on large, clear text as a foundational supervisor. By temporally back-tracking these high-confidence detections through video sequences, we automatically synthesize pseudo-labels for preceding frames where the distant text is still visually degraded or undersized. These pseudo-labels enable training a student model specialized for early text detection, without requiring any manual annotation. The success of this approach depends on accurate pseudo-label generation, for which we develop a dedicated scene text tracker capable of maintaining consistent text identities across challenging video sequences. In addition, we propose SceneText50, a diverse multilingual outdoor dataset to facilitate training and evaluation. Experiments show that our framework significantly improves early detection accuracy and robustness across varied scenes and languages. Code and data are at \href{https://github.com/trid2912/BackTrackingText}{https://github.com/trid2912/BackTrackingText}.