Title: Vision-based system identification and 3D keypoint discovery using dynamics constraints
Abstract: In this talk we'll introduce V-SysId, a novel method that enables simultaneous keypoint discovery, 3D system identification, and extrinsic camera calibration from an unlabeled video taken from a static camera, using only the family of equations of motion of the object of interest as weak supervision. V-SysId takes keypoint trajectory proposals and alternates between maximum likelihood parameter estimation and extrinsic camera calibration, before applying a suitable selection criterion to identify the track of interest. This is then used to train a keypoint tracking model using supervised learning. We will show and discuss applications to robotics, physics, and physiology settings to highlight the utility of this approach.
Title: SLAM in dynamic environments with large occlusion with the aid of robot proprioception
Abstract: In this talk, we will introduce RigidFusion, which proposes a novel pipeline to enable simultaneous localisation and mapping in dynamic environments with large occlusion. Large occlusion means that the dynamic objects can occupy the major components of the camera view while the state-of-the-art methods are unable to distinguish the static background from dynamic objects in this scenario. To solve this problem, RigidFusion uses motion priors from robot proprioception and simultaneously segment, track, and model both the static and dynamic components from RGB-D images. We also show RigidFusion’s experiment results, potential applications, and current limitations.