Tsukada Lab. Computer Network Group
← Publications Int'l conference

Enhancing Unsupervised Person Re-Identification with GAN-Driven Multi-Pose Augmentation

Romina Zakerian, Ehsan Javanmardi, Manabu Tsukada, Mahdi Javanmardi, Mohammad Rahmati

International Conference on Intelligent Computing and its Emerging Applications (ICEA 2024), Tokyo, Japan, pp. 27–32, Nov 28, 2024

Summary in three lines

  • The study tackles re-identification across camera views without manual identity labels.
  • GAN-generated images add new poses of the same person through four augmentation approaches.
  • The reported gains over baselines reach 4.1% in Rank-1 and 13.1% in mean average precision.

Abstract

Person re-identification is a crucial computer vision task, particularly for surveillance systems. This process aims to retrieve the identity of individuals across non-overlapping camera views. Despite significant advancements, the manual annotation process makes real-world deployments less scalable. To address this challenge, data augmentation techniques such as Generative Adversarial Networks (GANs) have proven effective in various studies. However, there is limited research combining unsupervised learning with GAN-based data augmentation for person re-identification. This study proposes leveraging unsupervised learning and GANs to enhance person re-identification. We generate images of the same person in a new pose using GANs and utilize these images across different independent methods to train the person re-identification model. The goal is to improve model robustness and generalization by exposing it to a wider range of poses and appearance variations. We explore four distinct methods to enhance model learning through data augmentation and evaluate their performance. The results demonstrate that our approach consistently outperforms the baseline methods, leading to an improvement of 4.1 percent in Rank-1 and 13.1 percent in mAP.

How to cite

R. Zakerian, E. Javanmardi, M. Tsukada, M. Javanmardi, and M. Rahmati, “Enhancing Unsupervised Person Re-Identification with GAN-Driven Multi-Pose Augmentation,” in International Conference on Intelligent Computing and its Emerging Applications (ICEA 2024), Tokyo, Japan, 2024, doi: 10.1145/3732437.3732742.

BibTeX

@inproceedings{Zakerian2024,
  title         = {Enhancing Unsupervised Person Re-Identification with GAN-Driven Multi-Pose Augmentation},
  author        = {Romina Zakerian and Ehsan Javanmardi and Manabu Tsukada and Mahdi Javanmardi and Mohammad Rahmati},
  doi           = {10.1145/3732437.3732742},
  year          = {2024},
  date          = {2024-11-28},
  urldate       = {2024-11-28},
  booktitle     = {International Conference on Intelligent Computing and its Emerging Applications (ICEA 2024)},
  pages         = {27--32},
  address       = {Tokyo, Japan},
  abstract      = {Person re-identification is a crucial computer vision task, particularly for surveillance systems. This process aims to retrieve the identity of individuals across non-overlapping camera views. Despite significant advancements, the manual annotation process makes real-world deployments less scalable. To address this challenge, data augmentation techniques such as Generative Adversarial Networks (GANs) have proven effective in various studies. However, there is limited research combining unsupervised learning with GAN-based data augmentation for person re-identification. This study proposes leveraging unsupervised learning and GANs to enhance person re-identification. We generate images of the same person in a new pose using GANs and utilize these images across different independent methods to train the person re-identification model. The goal is to improve model robustness and generalization by exposing it to a wider range of poses and appearance variations. We explore four distinct methods to enhance model learning through data augmentation and evaluate their performance. The results demonstrate that our approach consistently outperforms the baseline methods, leading to an improvement of 4.1 percent in Rank-1 and 13.1 percent in mAP.},
  keywords      = {ITS},
  pubstate      = {published},
  tppubtype     = {inproceedings}
}