Mingchao Liang

Multiobject tracking (MOT) is a key challenge in a wide range of applications, such as autonomous navigation and applied ocean sciences. MOT is complicated by object appearance and disappearance, data association ambiguities, and occlusion. Conventional Bayesian methods for MOT, e.g., methods based on belief propagation (BP), entirely rely on a statistical model. This fully model-based approach can lead to highly suboptimal estimates when there is a mismatch between the statistical model and the true data-generating process. In this webinar, the presenter will introduce a model-based and data-driven MOT method that relies on neural-enhanced belief propagation (NEBP) to integrate hand-designed statistical models with neural networks. In this way, the proposed hybrid MOT approach can combine the advantages of model-based and data-driven estimation, i.e., it makes use of information learned from raw sensor data but also relies on a scalable statistical model. NEBP for MOT is currently leading the nuScenes tracking challenge for LiDAR data.