Knowledge Distillation for Multi-Target Domain Adaptation in Real-Time Person Re-Identification

  • 2022-05-12 18:28:02
  • FĂ©lix Remigereau, Djebril Mekhazni, Sajjad Abdoli, Le Thanh Nguyen-Meidine, Rafael M. O. Cruz, Eric Granger
  • 1


Despite the recent success of deep learning architectures, personre-identification (ReID) remains a challenging problem in real-wordapplications. Several unsupervised single-target domain adaptation (STDA)methods have recently been proposed to limit the decline in ReID accuracycaused by the domain shift that typically occurs between source and targetvideo data. Given the multimodal nature of person ReID data (due to variationsacross camera viewpoints and capture conditions), training a common CNNbackbone to address domain shifts across multiple target domains, can providean efficient solution for real-time ReID applications. Although multi-targetdomain adaptation (MTDA) has not been widely addressed in the ReID literature,a straightforward approach consists in blending different target datasets, andperforming STDA on the mixture to train a common CNN. However, this approachmay lead to poor generalization, especially when blending a growing number ofdistinct target domains to train a smaller CNN. To alleviate this problem, we introduce a new MTDA method based on knowledgedistillation (KD-ReID) that is suitable for real-time person ReID applications.Our method adapts a common lightweight student backbone CNN over the targetdomains by alternatively distilling from multiple specialized teacher CNNs,each one adapted on data from a specific target domain. Extensive experimentsconducted on several challenging person ReID datasets indicate that ourapproach outperforms state-of-art methods for MTDA, including blending methods,particularly when training a compact CNN backbone like OSNet. Results suggestthat our flexible MTDA approach can be employed to design cost-effective ReIDsystems for real-time video surveillance applications.


Quick Read (beta)

loading the full paper ...