References & Citations
Computer Science > Computer Vision and Pattern Recognition
Title: Multi-modal In-Context Learning Makes an Ego-evolving Scene Text Recognizer
(Submitted on 22 Nov 2023 (v1), last revised 28 Mar 2024 (this version, v3))
Abstract: Scene text recognition (STR) in the wild frequently encounters challenges when coping with domain variations, font diversity, shape deformations, etc. A straightforward solution is performing model fine-tuning tailored to a specific scenario, but it is computationally intensive and requires multiple model copies for various scenarios. Recent studies indicate that large language models (LLMs) can learn from a few demonstration examples in a training-free manner, termed "In-Context Learning" (ICL). Nevertheless, applying LLMs as a text recognizer is unacceptably resource-consuming. Moreover, our pilot experiments on LLMs show that ICL fails in STR, mainly attributed to the insufficient incorporation of contextual information from diverse samples in the training stage. To this end, we introduce E$^2$STR, a STR model trained with context-rich scene text sequences, where the sequences are generated via our proposed in-context training strategy. E$^2$STR demonstrates that a regular-sized model is sufficient to achieve effective ICL capabilities in STR. Extensive experiments show that E$^2$STR exhibits remarkable training-free adaptation in various scenarios and outperforms even the fine-tuned state-of-the-art approaches on public benchmarks. The code is released at this https URL .
Submission history
From: Hao Liu [view email][v1] Wed, 22 Nov 2023 02:46:57 GMT (542kb,D)
[v2] Thu, 23 Nov 2023 08:15:18 GMT (716kb,D)
[v3] Thu, 28 Mar 2024 08:30:56 GMT (1082kb,D)
Link back to: arXiv, form interface, contact.