# Optimizing speed/accuracy trade-off for person re-identification via knowledge distillation

## TL;DR

This paper compares classical and deep learning methods for person re-identification, and introduces network distillation to improve speed-accuracy trade-offs, achieving better performance with reduced computational cost.

## Contribution

It proposes using knowledge distillation to enhance deep learning models for person re-identification, balancing accuracy and speed effectively.

## Key findings

- Distillation reduces inference computational cost.
- Distillation improves accuracy over baseline deep models.
- Deep models outperform classical methods in accuracy.

## Abstract

Finding a person across a camera network plays an important role in video surveillance. For a real-world person re-identification application, in order to guarantee an optimal time response, it is crucial to find the balance between accuracy and speed. We analyse this trade-off, comparing a classical method, that comprises hand-crafted feature description and metric learning, in particular, LOMO and XQDA, to deep learning based techniques, using image classification networks, ResNet and MobileNets. Additionally, we propose and analyse network distillation as a learning strategy to reduce the computational cost of the deep learning approach at test time. We evaluate both methods on the Market-1501 and DukeMTMC-reID large-scale datasets, showing that distillation helps reducing the computational cost at inference time while even increasing the accuracy performance.