Interactive Cloth Perception

Abstract: Deformable objects are common in our environments, yet, they possess a set of unique properties that make them incredibly difficult for robots to perceive and interact with -- they have near-infinite degrees of freedom and are subject to extreme cases of self-occlusion, especially when folded or crumpled. To simplify the problem, prior works on cloth manipulation often build on simplifying assumptions such as full visibility, or known instance-level meshes. In this talk, I will first introduce perception and manipulation algorithms that are able to generalize beyond these conditions (i.e., GarmentsNet and FlingBot), and then discuss how robots perception and interaction capability could be learned jointly in order to mutually benefit each other, moving towards an interactive perception framework.

Bio: Shuran Song is an Assistant Professor in the Department of Computer Science at Columbia University. Her research focuses on computer vision and robotics. Shuran received her PhD degree in Computer Science from Princeton University in 2018, and before that, completed a BEng. in Computer Science at HKUST in 2013. During her PhD, she spent time working at Microsoft Research and Google. Shuran has received numerous awards, including the Best Systems Paper Award at RSS 2019, IEEE Transactions on Robotics Best Paper Award in 2020, Amazon Research Award in 2020, and TRI Young Faculty Award in 2021. She is also a 2021 Microsoft Research Faculty Fellow. Shuran was part of the MIT-Princeton team for the Amazon Robotics Challenge, winning the 3rd place in 2016 and 1st place (stow task) in 2017. To learn more about Shuran’s work visit: https://www.cs.columbia.edu/~shurans/

Видео Interactive Cloth Perception канала CITRIS and the Banatao Institute