Learning Non-Metric Visual Similarity for Image Retrieval

TL;DR

This paper introduces neural network-based non-metric similarity functions for image retrieval, demonstrating improved performance over traditional metric distances by better modeling human visual perception.

Contribution

The work proposes a differentiable, end-to-end trainable neural network model for learning non-metric visual similarity functions, enhancing image retrieval accuracy.

Findings

Non-metric similarity networks outperform metric distances in image retrieval tasks.

The approach improves retrieval performance on standard datasets.

Neural networks effectively model complex visual similarities.

Abstract

Measuring visual similarity between two or more instances within a data distribution is a fundamental task in image retrieval. Theoretically, non-metric distances are able to generate a more complex and accurate similarity model than metric distances, provided that the non-linear data distribution is precisely captured by the system. In this work, we explore neural networks models for learning a non-metric similarity function for instance search. We argue that non-metric similarity functions based on neural networks can build a better model of human visual perception than standard metric distances. As our proposed similarity function is differentiable, we explore a real end-to-end trainable approach for image retrieval, i.e. we learn the weights from the input image pixels to the final similarity score. Experimental evaluation shows that non-metric similarity networks are able to learn visual similarities between images and improve performance on top of state-of-the-art image representations, boosting results in standard image retrieval datasets with respect standard metric distances.