Deep Metric Learning using Similarities from Nonlinear Rank Approximations

TL;DR

This paper introduces a novel deep metric learning approach that uses nonlinear rank approximations to improve image retrieval quality by focusing on the most influential feature vectors.

Contribution

It proposes a new loss function based on normalized rank approximations converted to similarities, enhancing the discrimination of similar and dissimilar image features.

Findings

Significant improvement over existing methods on multiple datasets

Better contraction of similar and dispersion of dissimilar samples

Effective for various embedding sizes

Abstract

In recent years, deep metric learning has achieved promising results in learning high dimensional semantic feature embeddings where the spatial relationships of the feature vectors match the visual similarities of the images. Similarity search for images is performed by determining the vectors with the smallest distances to a query vector. However, high retrieval quality does not depend on the actual distances of the feature vectors, but rather on the ranking order of the feature vectors from similar images. In this paper, we introduce a metric learning algorithm that focuses on identifying and modifying those feature vectors that most strongly affect the retrieval quality. We compute normalized approximated ranks and convert them to similarities by applying a nonlinear transfer function. These similarities are used in a newly proposed loss function that better contracts similar and disperses dissimilar samples. Experiments demonstrate significant improvement over existing deep feature embedding methods on the CUB-200-2011, Cars196, and Stanford Online Products data sets for all embedding sizes.