Consistent camera-invariant and noise-tolerant learning for unsupervised person re-identification
作者：
Highlights：
• Research focus on stable training with noise labels caused by camera variations and other factors.
• A consistent meta-learning strategy is used to overcome camera variations.
• A novel loss introduces additive margin into unsupervised person re-identification.
• A noise-tolerant loss enhances the robust ability of training.
• The method achieves comparable performance with the state-of-the-art methods.
摘要
•Research focus on stable training with noise labels caused by camera variations and other factors.•A consistent meta-learning strategy is used to overcome camera variations.•A novel loss introduces additive margin into unsupervised person re-identification.•A noise-tolerant loss enhances the robust ability of training.•The method achieves comparable performance with the state-of-the-art methods.
论文关键词：Person re-identification,Meta learning,Noise pseudo label,Camera variations
论文评审过程：Received 24 November 2021, Revised 7 April 2022, Accepted 15 April 2022, Available online 20 April 2022, Version of Record 29 April 2022.
论文官网地址：https://doi.org/10.1016/j.imavis.2022.104462
