Spatio-Temporal Association and Curve Kernel Networks (STACKNet) for Multi-cue Data Association

Target tracking involves several steps including detection, filtering and data association. In the proposed effort we propose to develop deep learning based technologies for data association to improve the overall tracking accuracy. Deep learning has shown significant gains in object detection in various domains. More recently, it has shown significant promise in data association and multi-target tracking. Comparative evaluation with existing approaches for data association will be conducted to show benefits of the proposed approach. Approved for Public Release |21-MDA-10789 (21 Apr 21)