Improving Cross-Modal Retrieval with Set of Diverse Embeddings

TL;DR

This paper introduces a novel set-based embedding method for cross-modal retrieval that uses a new similarity function and a set prediction module to better capture diverse semantics in images and captions.

Contribution

It proposes a new similarity function called smooth-Chamfer similarity and a set prediction module using slot attention, advancing set-based embedding for cross-modal retrieval.

Findings

Outperforms existing methods on COCO and Flickr30K datasets.

Requires less computation at inference compared to some prior methods.

Effectively captures diverse semantics in image-caption pairs.

Abstract

Cross-modal retrieval across image and text modalities is a challenging task due to its inherent ambiguity: An image often exhibits various situations, and a caption can be coupled with diverse images. Set-based embedding has been studied as a solution to this problem. It seeks to encode a sample into a set of different embedding vectors that capture different semantics of the sample. In this paper, we present a novel set-based embedding method, which is distinct from previous work in two aspects. First, we present a new similarity function called smooth-Chamfer similarity, which is designed to alleviate the side effects of existing similarity functions for set-based embedding. Second, we propose a novel set prediction module to produce a set of embedding vectors that effectively captures diverse semantics of input by the slot attention mechanism. Our method is evaluated on the COCO and Flickr30K datasets across different visual backbones, where it outperforms existing methods including ones that demand substantially larger computation at inference.