Extending CLIP for Category-to-image Retrieval in E-commerce

TL;DR

This paper introduces CLIP-ITA, a multimodal model for category-to-image retrieval in e-commerce, effectively combining textual, visual, and attribute data to improve search accuracy.

Contribution

The paper proposes a novel multimodal model, CLIP-ITA, specifically designed for category-to-image retrieval in e-commerce, leveraging multiple data modalities for enhanced performance.

Findings

CLIP-ITA outperforms visual-only models in retrieval tasks.

Adding attribute information improves model accuracy.

Multimodal integration enhances e-commerce search effectiveness.

Abstract

E-commerce provides rich multimodal data that is barely leveraged in practice. One aspect of this data is a category tree that is being used in search and recommendation. However, in practice, during a user's session there is often a mismatch between a textual and a visual representation of a given category. Motivated by the problem, we introduce the task of category-to-image retrieval in e-commerce and propose a model for the task, CLIP-ITA. The model leverages information from multiple modalities (textual, visual, and attribute modality) to create product representations. We explore how adding information from multiple modalities (textual, visual, and attribute modality) impacts the model's performance. In particular, we observe that CLIP-ITA significantly outperforms a comparable model that leverages only the visual modality and a comparable model that leverages the visual and attribute modality.