Object Re-identification via Spatial-temporal Fusion Networks and Causal Identity Matching

TL;DR

This paper presents a novel object re-identification framework that combines spatial-temporal fusion networks with causal identity matching to improve accuracy in real-world camera networks, outperforming existing methods across multiple datasets.

Contribution

Introduces a new ReID framework integrating spatial-temporal cues with causal identity matching, addressing appearance similarity and real-world scenario challenges.

Findings

Achieves up to 99.70% rank-1 accuracy on VeRi776 dataset.

Demonstrates significant improvement in mAP and F1 scores with the proposed methods.

Effective across different data domains like vehicles and pedestrians.

Abstract

Object re-identification (ReID) in large camera networks faces numerous challenges. First, the similar appearances of objects degrade ReID performance, a challenge that needs to be addressed by existing appearance-based ReID methods. Second, most ReID studies are performed in laboratory settings and do not consider real-world scenarios. To overcome these challenges, we introduce a novel ReID framework that leverages a spatial-temporal fusion network and causal identity matching (CIM). Our framework estimates camera network topology using a proposed adaptive Parzen window and combines appearance features with spatial-temporal cues within the fusion network. This approach has demonstrated outstanding performance across several datasets, including VeRi776, Vehicle-3I, and Market-1501, achieving up to 99.70% rank-1 accuracy and 95.5% mAP. Furthermore, the proposed CIM approach, which dynamically assigns gallery sets based on camera network topology, has further improved ReID accuracy and robustness in real-world settings, evidenced by a 94.95% mAP and a 95.19% F1 score on the Vehicle-3I dataset. The experimental results support the effectiveness of incorporating spatial-temporal information and CIM for real-world ReID scenarios, regardless of the data domain (e.g., vehicle, person).