aiqbench analysis
This paper introduces GOT-JEPA, a novel framework that leverages Joint-Embedding Predictive Architecture to improve object tracking robustness against occlusions and scene changes. By training models to predict pseudo-tracking outputs from corrupted frames, the approach enhances generalization in dynamic environments without relying on specific target optimization. The accompanying OccuSolver module further refines visibility estimation through iterative point-centric adaptation, offering a significant step forward for reliable visual perception systems.
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 model-predictive pretraining framework that extends JEPA from predicting image features to predicting tracking models. Given identical historical information, a teacher predictor generates pseudo-tracking models from a clean current frame, and a student predictor learns to predict the same pseudo-tracking models from a corrupted version of the current frame. This design provides stable pseudo supervision and explicitly trains the predictor to produce reliable tracking models under occlusions, distractors, and other adverse observations, improving generalization to dynamic environments. Building on GOT-JEPA, we further propose OccuSolver to enhance occlusion perception for object tracking. OccuSolver adapts a point-centric point tracker for object-aware visibility estimation and detailed occlusion-pattern capture. Conditioned on object priors iteratively generated by the tracker, OccuSolver incrementally refines visibility states, strengthens occlusion handling, and produces higher-quality reference labels that progressively improve subsequent model predictions. Extensive evaluations on seven benchmarks show that our method effectively enhances tracker generalization and robustness.
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 model-predictive pretraining framework that extends JEPA from predicting image features to predicting tracking models. Given identical historical information, a teacher predictor generates pseudo-tracking models from a clean current frame, and a student predictor learns to predict the same pseudo-tracking models from a corrupted version of the current frame. This design provides stable pseudo supervision and explicitly trains the predictor to produce reliable tracking models under occlusions, distractors, and other adverse observations, improving generalization to dynamic environments. Building on GOT-JEPA, we further propose OccuSolver to enhance occlusion perception for object tracking. OccuSolver adapts a point-centric point tracker for object-aware visibility estimation and detailed occlusion-pattern capture. Conditioned on object priors iteratively generated by the tracker, OccuSolver incrementally refines visibility states, strengthens occlusion handling, and produces higher-quality reference labels that progressively improve subsequent model predictions. Extensive evaluations on seven benchmarks show that our method effectively enhances tracker generalization and robustness.
Accepted by IEEE Transactions on Circuits and Systems for Video Technology (TCSVT). This research focuses on learning model adaptation for adverse and dynamic environments, as well as fine-grained occlusion perception for tracking
Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Multimedia (cs.MM); Neural and Evolutionary Computing (cs.NE)
I.4; I.2; I.5; I.4.1; I.4.8; I.4.9; I.4.10; K.4; C.5; H.3; J.0
arXiv:2602.14771 [cs.CV]
(or arXiv:2602.14771v5 [cs.CV] for this version)
https://doi.org/10.1109/TCSVT.2026.3675005 DOI(s) linking to related resources
Full text under the publisher's open licence · 295 words
Related coverage
Improving CoT Monitorability of Evaluation Awareness via Verbalization Training
LessWrong · Oct 2, 2026
Scale AV Perception Across Vehicle Platforms with NVIDIA Omniverse NuRec
NVIDIA · Aug 31, 2026
[AINews] Reality Checks on AI News (Yegge shuts down Gas Town, Databricks’ +60% Astra cost)
Latent Space · Sep 17, 2026
Towards Accurate State Estimation: Motion Dynamics Kalman Filter for 3D Multi-Object Tracking
arXiv · Oct 14, 2026