3DMODT: Attention-Guided Affinities for Joint Detection & Tracking in 3D Point Clouds
Jyoti Kini,Ajmal Mian,Mubarak Shah,Jyoti Kini,Ajmal Mian,Mubarak Shah
We propose a method for joint detection and tracking of multiple objects in 3D point clouds, a task conventionally treated as a two-step process comprising object detection followed by data association. Our method embeds both steps into a single end-to-end trainable network eliminating the dependency on external object detectors. Our model exploits temporal information employing multiple frames to...
