Recently, perception algorithms based on deep learning have been on the rise, and these techniques are being applied to robotic manipulation tasks with promising results. In these manipulation tasks, perception is important to know the location of every object in order to plan and control for desired trajectories. This project focuses on using RGB-D images to best represent the state of the world for robotic manipulation. Our work plans to automatically detect key points that can be used for pose estimation and tracking. We aim to achieve results on both real-world data and meaningful tasks.

I’m participating in SuperUROP because I’m eager to gain research experience as I prepare for graduate school. After doing robotics UROPs in my freshman and sophomore years, I’m excited to commit to this longer project with hopes of contributing to the fields of computer vision and robotics.