Positive Pair Distillation Considered Harmful: Continual Meta Metric Learning for Lifelong Object Re-Identification
This work addresses the problem of incremental learning for object re-identification, which is crucial for applications like surveillance and robotics, but it is incremental as it builds on existing continual meta metric learning approaches.
The paper tackles the challenge of lifelong object re-identification by proposing a method that removes positive pairs from knowledge distillation to prevent forgetting and improve generalization to unseen identities, achieving significant state-of-the-art performance in intra-domain and inter-domain experiments.
Lifelong object re-identification incrementally learns from a stream of re-identification tasks. The objective is to learn a representation that can be applied to all tasks and that generalizes to previously unseen re-identification tasks. The main challenge is that at inference time the representation must generalize to previously unseen identities. To address this problem, we apply continual meta metric learning to lifelong object re-identification. To prevent forgetting of previous tasks, we use knowledge distillation and explore the roles of positive and negative pairs. Based on our observation that the distillation and metric losses are antagonistic, we propose to remove positive pairs from distillation to robustify model updates. Our method, called Distillation without Positive Pairs (DwoPP), is evaluated on extensive intra-domain experiments on person and vehicle re-identification datasets, as well as inter-domain experiments on the LReID benchmark. Our experiments demonstrate that DwoPP significantly outperforms the state-of-the-art. The code is here: https://github.com/wangkai930418/DwoPP_code