profile photo

Yuzhe Qin

Email: y1qin [at] ucsd (dot) edu

I am a second year master at UC San Diego, advised by Prof. Hao Su in the Center for Visual Computing. Before that, I received my bachelor degree from Shanghai Jiao Tong University.

Currently, I am a AI resident at X for summer 2020

CV /  Github /  Google Scholar

Research Interest

My research interests lie in building robot that can learn from interacting with the world. In particular, I am interested in robotics manipulation, task-oriented 3D perception and reinforcement learning. I am also interested in building simulation environment to train robot learning algorithm.

Publications
SAPIEN: A SimulAted Part-based Interactive ENvironment
Fanbo Xiang, Yuzhe Qin, Kaichun Mo, Yikuan Xia, Hao Zhu, Fangchen Liu, Minghua Liu, Hanxiao Jiang, Yifu Yuan, He Wang, Li Yi, Angel X.Chang, Leonidas J. Guibas and Hao Su
Conference on Computer Vision and Pattern Recognition (CVPR) 2020 (Oral)
Paper Project

We developed SAPIEN, a robotics simulator for training robot by interacting with articulated objects. Our realistic and physics-rich simulated environment hosts a large-scale set for articulated objects. It also provided a hierarchical robot interface consists of sensors and controllers to support both low-level control and high-level planning.

S4G: Amodal Single-view Single-Shot SE(3) Grasp Detection in Cluttered Scenes
Yuzhe Qin*, Rui Chen*, Hao Zhu, Meng Song, Jing Xu, Hao Su
Conference of Robot Learning (CoRL) 2019
Paper Project

We studied the problem of 6D grasping in cluttered scene captured with commodity depth sensor from a single view point. Our learning based approach trained in a synthetic scene can work well in real-world scenarios, with improved speed and success rate compared with SOTA.

Composing Task-Agnostic Policies with Deep Reinforcement Learning
Ahmed Qureshi, Jacob Johnson, Yuzhe Qin, Taylor Henderson, Byron Boots, Michael Yip
International Conference of Learning Representation (ICLR) 2020
Paper Project

We propose a novel reinforcement learning-based skill transfer and composition method that takes the agent's primitive policies to solve unseen tasks. We show that our method not only transfers skills to new problem settings but also solves the challenging environments requiring both task planning and motion control with high data efficiency.

Honors and Awards
  • Zhiyuan Honor Degree of B.Sc (48 among 3992), Shanghai Jiao Tong University. 2018
  • Zhiyuan College Honors Scholarship (Top 5%), Shanghai Jiao Tong University. 2015 & 2016
  • Excellecnt Graduate Award, Shanghai. 2018

Website source from Jon Barron