提出DAAL损失函数，提升多模态特征的类内分布与类间分离。
DAAL: Density-Aware Adaptive Line Margin Loss for Multi-Modal Deep Metric Learning
- 基于自适应线性策略，动态调整类内子簇边界。
- 在细粒度数据集上显著提升检索性能，优于传统方法。
- 适合需要精准特征表示的图像检索与识别任务。
多模态深度度量学习在人脸识别、细粒度物体识别和商品搜索等任务中至关重要。传统度量学习方法（基于距离或间隔）主要关注类间分离，常忽略对多模态特征学习至关重要的类内分布。为此，本文提出一种新型损失函数——密度感知自适应线边距损失（DAAL），在保持嵌入密度分布的同时，促进每类内部自适应子簇的形成。通过自适应线策略，DAAL不仅增强类内差异性，还确保强类间分离，从而实现有效的多模态表征。在多个基准细粒度数据集上的全面实验表明，DAAL性能优越，展现出在检索应用和多模态深度度量学习中的巨大潜力。
原文摘要 · Abstract (English)
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.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。