AI Research Activities in the Intelligent Machine Perception Group in 2021
The Intelligent Machine Perception group focuses on fundamental and applied research in computer vision, machine learning and robotics. The group is open to research collaboration in the fields of robust object pose estimation, human motion recognition, or robot motion learning. Examples include: (i) verifying our state-of-the-art methods on real-world industrial data or (ii) finding new challenging tasks from the industrial domain. We present here the key results that are becoming mature enough to be used in real-world industrial setups.
Robust Object Pose Estimation from Images
CosyPose
Winner of the BOP Challenge 2020Yann Labbé, Justin Carpentier, Mathieu Aubry, Josef Sivic
ECCV: European Conference on Computer Vision, 2020
6D object pose estimation optimizing multi-view COnSistencY. Given a set of RGB images depicting a scene with known objects taken from unknown viewpoints, our method accurately reconstructs the scene, recovering all objects in the scene, their 6D pose and the camera viewpoints.
Human-Object Motion and Force Estimation
Zongmian Li, Jiri Sedlar, Justin Carpentier, Ivan Laptev, Nicolas Mansard, Josef Sivic
Estimating 3D Motion and Forces of Human-Object Interactions from Internet Videos.
Learning to Manipulate Tools by Watching Video
Learning to Manipulate Tools by Aligning Simulation to Video Demonstration
K. Zorina, J. Carpentier, J. Sivic, V. Petrík
A seamless integration of robots into human environments requires robots to learn how to use existing human tools. Current approaches for learning tool manipulation skills mostly rely on expert demonstrations provided in the target robot environment, for example, by manually guiding the robot manipulator or by teleoperation. In this work, we introduce an automated approach that replaces an expert demonstration with a Youtube video for learning a tool manipulation strategy.