Attention-based Ensemble for Deep Metric Learning

TL;DR

This paper introduces an attention-based ensemble method with divergence loss for deep metric learning, significantly improving image retrieval performance by promoting diverse feature attention among learners.

Contribution

It proposes a novel attention-based ensemble framework with a divergence loss to enhance diversity and performance in deep metric learning.

Findings

Outperforms state-of-the-art methods on standard benchmarks

Achieves significant improvements in image retrieval accuracy

Demonstrates the effectiveness of attention masks and divergence loss

Abstract

Deep metric learning aims to learn an embedding function, modeled as deep neural network. This embedding function usually puts semantically similar images close while dissimilar images far from each other in the learned embedding space. Recently, ensemble has been applied to deep metric learning to yield state-of-the-art results. As one important aspect of ensemble, the learners should be diverse in their feature embeddings. To this end, we propose an attention-based ensemble, which uses multiple attention masks, so that each learner can attend to different parts of the object. We also propose a divergence loss, which encourages diversity among the learners. The proposed method is applied to the standard benchmarks of deep metric learning and experimental results show that it outperforms the state-of-the-art methods by a significant margin on image retrieval tasks.