GOT-JEPA: Generic Object Tracking with Model Adaptation and Occlusion Handling using Joint-Embedding Predictive Architecture
arXiv:2602.14771v5 Announce Type: replace-cross Abstract: The human visual system tracks objects by integrating current observations with previously observed information, adapting to target and scene changes, and reasoning about occlusion at fine granularity. In contrast, recent generic object trackers are often optimized for training targets, which limits robustness and generalization in unseen scenarios, and their occlusion reasoning remains coarse, lacking detailed modeling of occlusion patterns. To address these limitations in generalization and occlusion perception, we propose GOT-JEPA, a
The continuous advancements in AI, particularly in computer vision and predictive modeling, enable new approaches to complex problems like robust object tracking in real-world environments.
Improved generic object tracking with advanced occlusion handling is crucial for highly reliable autonomous systems, advanced surveillance, and human-robot interaction in unpredictable settings.
Current limitations in tracking robustness and generalization, especially under occlusion, are being addressed, leading to more resilient AI applications in autonomous perception.
- · Autonomous vehicle developers
- · Robotics industry
- · Surveillance technology providers
- · AI hardware manufacturers
- · Developers of less robust tracking algorithms
- · Systems highly reliant on pristine visual inputs
More reliable deployment of AI agents and autonomous systems in complex, dynamic visual environments.
Accelerated development and adoption of AI-powered applications that require persistent object understanding.
Enhanced human-machine collaboration due to AI systems having a more sophisticated understanding of objects and their interactions within shared spaces.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI