Embedding Compression for Efficient Re-Identification

TL;DR

This paper explores various compression techniques for re-identification embeddings, demonstrating that they can be significantly reduced in size with minimal performance loss, thus addressing scalability issues.

Contribution

It benchmarks multiple compression methods for ReID embeddings, revealing that up to 96x compression is achievable with little impact on accuracy.

Findings

ReID embeddings can be compressed by up to 96x.

Compression causes minimal performance drop.

High-dimensional space is underutilized in current ReID systems.

Abstract

Real world re-identfication (ReID) algorithms aim to map new observations of an object to previously recorded instances. These systems are often constrained by quantity and size of the stored embeddings. To combat this scaling problem, we attempt to shrink the size of these vectors by using a variety of compression techniques. In this paper, we benchmark quantization-aware-training along with three different dimension reduction methods: iterative structured pruning, slicing the embeddings at initialize, and using low rank embeddings. We find that ReID embeddings can be compressed by up to 96x with minimal drop in performance. This implies that modern re-identification paradigms do not fully leverage the high dimensional latent space, opening up further research to increase the capabilities of these systems.