Faster Person Re-Identification (2008.06826v1)

Abstract: Fast person re-identification (ReID) aims to search person images quickly and accurately. The main idea of recent fast ReID methods is the hashing algorithm, which learns compact binary codes and performs fast Hamming distance and counting sort. However, a very long code is needed for high accuracy (e.g. 2048), which compromises search speed. In this work, we introduce a new solution for fast ReID by formulating a novel Coarse-to-Fine (CtF) hashing code search strategy, which complementarily uses short and long codes, achieving both faster speed and better accuracy. It uses shorter codes to coarsely rank broad matching similarities and longer codes to refine only a few top candidates for more accurate instance ReID. Specifically, we design an All-in-One (AiO) framework together with a Distance Threshold Optimization (DTO) algorithm. In AiO, we simultaneously learn and enhance multiple codes of different lengths in a single model. It learns multiple codes in a pyramid structure, and encourage shorter codes to mimic longer codes by self-distillation. DTO solves a complex threshold search problem by a simple optimization process, and the balance between accuracy and speed is easily controlled by a single parameter. It formulates the optimization target as a $F_{\beta}$ score that can be optimised by Gaussian cumulative distribution functions. Experimental results on 2 datasets show that our proposed method (CtF) is not only 8% more accurate but also 5x faster than contemporary hashing ReID methods. Compared with non-hashing ReID methods, CtF is $50\times$ faster with comparable accuracy. Code is available at https://github.com/wangguanan/light-reid.

• The paper presents a novel Coarse-to-Fine (CtF) hashing strategy that accelerates person re-identification by using short binary codes for preliminary ranking and long codes for precise refinement.
• The work integrates an All-in-One (AiO) framework with self-distillation to optimize multiple code lengths concurrently, enhancing the efficiency of binary code learning.
• Experimental results demonstrate an 8% accuracy boost and up to 50x faster performance compared to traditional methods, highlighting its practical impact for time-sensitive applications.

Faster Person Re-Identification: A Coarse-to-Fine Hashing Approach

The paper "Faster Person Re-Identification" by Wang et al. addresses the challenge of enhancing the speed and accuracy of person re-identification (ReID) systems through improved hashing techniques. The authors introduce a novel Coarse-to-Fine (CtF) search strategy that leverages both short and long binary codes to optimize performance.

Key Contributions

1. Coarse-to-Fine (CtF) Strategy: The core innovation is the CtF hashing approach, which uses shorter binary codes for preliminary ranking and longer codes for refining top candidates. This methodology aims to balance speed with accuracy by minimizing the computational overhead associated with evaluating a large gallery.
2. All-in-One (AiO) Framework: The AiO framework facilitates the simultaneous learning and enhancement of multiple code lengths via self-distillation. It employs a pyramid structure where shorter codes are fine-tuned to approximate the performance of longer codes.
3. Distance Threshold Optimization (DTO): To determine optimal distance thresholds for CtF, the DTO algorithm utilizes a Gaussian cumulative distribution to optimize the $F_{\beta}$ score, ensuring an effective trade-off between speed and accuracy.

Numerical Results

The authors report that their CtF method outperforms traditional hashing methods, achieving an 8% accuracy improvement and a fivefold increase in speed. Compared to non-hashing ReID methods, CtF is 50 times faster while maintaining competitive levels of accuracy. These claims are substantiated through extensive experiments on datasets such as Market-1501 and DukeMTMC-reID.

Practical and Theoretical Implications

The proposed CtF strategy significantly accelerates ReID processes, making it highly applicable in time-sensitive domains like surveillance and security. The ability to effectively utilize both short and long codes within a single framework suggests potential applications in other fine-grained search tasks beyond ReID.

Theoretically, the integration of self-distillation within the AiO framework offers insights into improving the learning efficiency of hashing methods. This approach not only enhances the performance of binary codes but also contributes to the broader understanding of multi-stage ranking processes in machine learning models.

Future Directions

Future research may explore extending the CtF and AiO frameworks to other domains requiring rapid search capabilities, such as image or document retrieval. Additionally, investigating the scalability of these methods in even larger datasets could provide valuable insights into their robustness and adaptability in real-world settings.

Conclusion

Wang et al.'s work presents an efficient approach to person ReID through the innovative use of hashing and search strategies. The CtF method exemplifies a significant advancement in balancing the trade-off between accuracy and speed, setting a new benchmark for future research in fast person re-identification systems.