.png){kind=icon}
.png){kind=icon}
Return
PLA: progressive learning algorithm for efficient person re-identification
.png){kind=icon}
PRE
AI
DOI：10.1007/s11042-022-12022-y.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Inthis paper, we study the problem of Person Re-Identification (ReID) for large-scale applications in the real-world scenarios. Recently most research efforts on ReID have been mainly devoted to building complicated part models, which however introduce considerably high computational cost and memory consumption, inhibiting its practicability in large-scale applications in practice. This paper aims to develop a novel learning strategy to find efficient feature embeddings while maintaining the balance of accuracy and model complexity. More specifically, we find by enhancing the classical triplet loss together with cross-entropy loss, our method can explore the hard examples and build a discriminant feature embedding yet compact enough for large-scale applications. Our training process is carried out progressively using Bayesian optimization, and we call it the Progressive Learning Algorithm (PLA). Extensive experiments on three large-scale datasets show that our PLA is comparable or better than the-state-of-the-arts. In particular, on the challenging Market-1501 dataset, we achieve Rank-1 = 94.7%/mAP= 89.4% while saving at least 30% parameters than strong part models. Finally, extra experimental results indicate that current neural network backbones can benefit from our PLA with an average performance improvement of approximately 2.23% and 1.63% regarding mAP and Rank-1, respectively.
Keywords:
ReID
Progressive learning
Bayesian optimization
Computational efficiency
Journal
IF:
3
Papers:
1.9W
Citations:
3.2W