[PaperReview] Unsupervised Cross-domain Image Retrieval via Prototypical Optimal Transport

 

Li, Bin, et al. “Unsupervised cross-domain image retrieval via prototypical optimal transport.” Proceedings of the AAAI Conference on Artificial Intelligence. Vol. 38. No. 4. 2024.

Paper Link: AAAI 2024

Introduction

This work introduces prototypical optimal transport for unsupervised cross-domain image retrieval. The method leverages optimal transport theory to align domains using prototypical representations.

Key Contributions

  • Prototypical optimal transport framework
  • Domain alignment using optimal transport
  • Improved cross-domain retrieval performance

Methodology

The approach uses prototypical representations and optimal transport to align feature distributions across domains, enabling effective cross-domain retrieval without paired data.

Figures

Note: This paper was published at AAAI 2024. Images will be added once the arXiv version is available or from the conference proceedings.

Results

Experimental results demonstrate the effectiveness of prototypical optimal transport in cross-domain image retrieval tasks.