Semantic Relatedness Based Re-ranker for Text Spotting

TL;DR

This paper introduces a neural relatedness model to enhance text spotting in images by leveraging semantic context, significantly improving recognition accuracy over traditional similarity measures.

Contribution

It proposes a novel neural approach to learn semantic relatedness tailored for text spotting tasks, addressing limitations of existing similarity measures.

Findings

Improved text spotting performance by up to 2.9 points.

Outperformed existing similarity measures on benchmark datasets.

Demonstrated the importance of semantic relatedness in vision-language tasks.

Abstract

Applications such as textual entailment, plagiarism detection or document clustering rely on the notion of semantic similarity, and are usually approached with dimension reduction techniques like LDA or with embedding-based neural approaches. We present a scenario where semantic similarity is not enough, and we devise a neural approach to learn semantic relatedness. The scenario is text spotting in the wild, where a text in an image (e.g. street sign, advertisement or bus destination) must be identified and recognized. Our goal is to improve the performance of vision systems by leveraging semantic information. Our rationale is that the text to be spotted is often related to the image context in which it appears (word pairs such as Delta-airplane, or quarters-parking are not similar, but are clearly related). We show how learning a word-to-word or word-to-sentence relatedness score can improve the performance of text spotting systems up to 2.9 points, outperforming other measures in a benchmark dataset.