Abstract:The objective of deep metric learning (DML) is to learn embeddings that can capture semantic similarity and dissimilarity information among data points. Existing pairwise or tripletwise loss functions used in DML are known to suffer from slow convergence due to a large proportion of trivial pairs or triplets as the model improves. To improve this, ranking-motivated structured losses are proposed recently to incorporate multiple examples and exploit the structured information among them. They converge faster and achieve state-of-the-art performance. In this work, we unveil two limitations of existing ranking-motivated structured losses and propose a novel ranked list loss to solve both of them. First, given a query, only a fraction of data points is incorporated to build the similarity structure. Consequently, some useful examples are ignored and the structure is less informative. To address this, we propose to build a set-based similarity structure by exploiting all instances in the gallery. The learning setting can be interpreted as few-shot retrieval: given a mini-batch, every example is iteratively used as a query, and the rest ones compose the gallery to search, i.e., the support set in few-shot setting. The rest examples are split into a positive set and a negative set. For every mini-batch, the learning objective of ranked list loss is to make the query closer to the positive set than to the negative set by a margin. Second, previous methods aim to pull positive pairs as close as possible in the embedding space. As a result, the intraclass data distribution tends to be extremely compressed. In contrast, we propose to learn a hypersphere for each class in order to preserve useful similarity structure inside it, which functions as regularisation. Extensive experiments demonstrate the superiority of our proposal by comparing with the state-of-the-art methods on the fine-grained image retrieval task. Our source code is available online: https://github.com/XinshaoAmosWang/Ranked-List-Loss-for-DML.

Online Progressive Deep Metric Learning

A Multilayer Framework for Online Metric Learning

Online Discriminative Dictionary Learning for Image Classification Based on Block-Coordinate Descent Method

Markov-Lipschitz Deep Learning

Self-distillation with Online Diffusion on Batch Manifolds Improves Deep Metric Learning

Improving Deep Metric Learning Via Self-Distillation and Online Batch Diffusion Process

MODL: Multilearner Online Deep Learning

Evolving Metric Learning for Incremental and Decremental Features

Ranked List Loss for Deep Metric Learning

OPML: A One-Pass Closed-Form Solution for Online Metric Learning

Towards Self-Adaptive Metric Learning On the Fly

Lifelong Metric Learning

Deep Factorized Metric Learning

Deep Metric Learning Based on Meta-Mining Strategy With Semiglobal Information

Deep Metric Learning Via Adaptive Learnable Assessment

Deep Localized Metric Learning.

Deep Adversarial Metric Learning

Exponential Discriminative Metric Embedding in Deep Learning

Deep Variational Metric Learning