Don't only Feel Read: Using Scene text to understand advertisements

TL;DR

This paper presents a framework that enhances advertisement image classification by integrating visual features with textual cues extracted from embedded scene text, demonstrating improved semantic understanding.

Contribution

It introduces a novel approach combining visual and textual features for ad classification, emphasizing the importance of scene text in semantic interpretation.

Findings

Textual cues improve classification accuracy

Scene text provides meaningful semantic information

Framework outperforms visual-only methods

Abstract

We propose a framework for automated classification of Advertisement Images, using not just Visual features but also Textual cues extracted from embedded text. Our approach takes inspiration from the assumption that Ad images contain meaningful textual content, that can provide discriminative semantic interpretetion, and can thus aid in classifcation tasks. To this end, we develop a framework using off-the-shelf components, and demonstrate the effectiveness of Textual cues in semantic Classfication tasks.