Технический репорт Almaz Zinollayev (BTSDigital) на CVPR:
sites.google.com/view/cvpr20-scalability/wod-reports

3D TRACKING (3rd place)
Given a temporal sequence (prior frames) of lidar and camera data, produce a set of 3D upright boxes and the correspondences between boxes across frames. (Entries that use future frames will be ineligible for awards.) You may use the baseline model or object detections linked below, if you wish.

drive.google.com/file/d/1DmrCFHO9hYfYn-Pp4UkO7wikuc4N0kWh/view