Multi view tracking improves online identity matching without prior training
OmniGlimpse: Asymmetric Identity Memory for Online Multi-View Tracking
Computer Vision and Pattern Recognition
Summary
Tracking people or objects using multiple cameras can be tricky because each view might see them differently or lose sight temporarily. The authors work on a way to keep track of identities across cameras live, without needing special preparation like training on that scene or calibrating cameras. They built a system called OmniGlimpse, which cleverly proposes guesses about who is who using short-term clues and verifies those guesses with long-term stable information. Their approach adapts quickly to changes but also remembers reliable identity traits, leading to much better tracking results than previous online methods.
What this means in practice
- •For security monitoring teams: Enable real-time reliable tracking of people across multiple surveillance cameras in changing environments without needing pre-calibrated setups.
- •For robotics developers: Allow robots with multiple cameras to maintain consistent identification of objects or people during navigation without scene-specific training.
Authors
Rangya Zhang, Jiaping Xiao, Yujing Shang, Yuhang Zhang, Mingsheng Li, Mir Feroskhan
Abstract
Multi-view tracking extends visual coverage beyond a single camera, providing complementary observations when targets are occluded or leave the view. This raises a question: how can observations from different views be matched to the correct identities without offline scene-specific training or camera calibration? Progress toward this goal remains limited because many offline formulations assume such deployment-specific knowledge, while online adaptation can amplify association errors. We study this problem as online multi-view identity maintenance (OMIM): targets are initialized once per view, with no further identity labels provided after each initialization. To this end, we introduce OmniGlimpse, an asymmetric propose-verify framework that separates adaptive association from stable identity consolidation. At the representation level, short-term local and cross-view evidence is mapped into target-specific adaptive spaces to propose associations, whereas long-term identity distributions are maintained for verification in a canonical space defined by a frozen encoder. At the update level, short-term evidence may guide current associations, whereas, after bootstrapping, only verified observations may update persistent memory or drive adaptation. Under a unified target-specified evaluation protocol, OmniGlimpse improves over the best evaluated online baseline for each metric by 9.1 IDF1 and 22.1 CVIDF1 points on M3Track. The corresponding gains on MvMHAT are 20.3 and 26.2 points. Despite using no offline scene-specific training, it outperforms offline-trained references in IDF1 and remains competitive in cross-view association. These results support an approach to online identity maintenance that adapts to changing appearances while consolidating reliable evidence into stable identity anchors.