DAAL: Density-Aware Adaptive Line Margin Loss for Multi-Modal Deep Metric Learning

TL;DR

DAAL introduces a novel density-aware adaptive margin loss that improves multi-modal deep metric learning by preserving intra-class density distributions and forming adaptive sub-clusters, leading to better retrieval performance.

Contribution

The paper proposes DAAL, a new loss function that maintains intra-class density and enables adaptive sub-cluster formation, advancing multi-modal deep metric learning.

Findings

DAAL outperforms existing methods on benchmark fine-grained datasets.

It enhances intra-class variance while maintaining inter-class separation.

Experiments demonstrate improved retrieval accuracy with DAAL.

Abstract

Multi-modal deep metric learning is crucial for effectively capturing diverse representations in tasks such as face verification, fine-grained object recognition, and product search. Traditional approaches to metric learning, whether based on distance or margin metrics, primarily emphasize class separation, often overlooking the intra-class distribution essential for multi-modal feature learning. In this context, we propose a novel loss function called Density-Aware Adaptive Margin Loss(DAAL), which preserves the density distribution of embeddings while encouraging the formation of adaptive sub-clusters within each class. By employing an adaptive line strategy, DAAL not only enhances intra-class variance but also ensures robust inter-class separation, facilitating effective multi-modal representation. Comprehensive experiments on benchmark fine-grained datasets demonstrate the superior performance of DAAL, underscoring its potential in advancing retrieval applications and multi-modal deep metric learning.