Pang-Chi (Sean) Lo

I am a first-year Master of Science in Computer Vision student at Carnegie Mellon University. Currently, I am working under Prof. Katerina Fragkiadaki's research group. I completed my Bachelor of Science in Computer Science and Information Engineering degree at National Taiwan University. During my time at NTU, I had the privilege to be advised by Prof. Winston H. Hsu, Prof. Yi-Ting Chen, and Prof. Tsung-Wei Ke. My primary research interests lie in robot learning, robotic perception, and 4D generative models

Email  /  Google Scholar  /  Github  /  LinkedIn  /  CV

Research

While existing feed-forward approaches typically address geometry reconstruction and motion estimation separately and often predict motions in mixture of camera movement, SM4RT solves these tasks sequentially by first reconstructing 3D geometry and then bootstrapping motion estimation from the predicted structure.

Existing Vision-Instruction Correlation (VIC) reward models struggle with training for long-horizon tasks. We propose VICtoR, a new reward model for long-horizon robotic reinforcement learning that assigns rewards hierarchically.

The novel adaptable error detection (AED) problem is formulated for monitoring few-shot imitation policies' behaviors, and we propose PrObe to address the challenging problem by learning from the policy's feature representations.

Professional Experience

Selected in Talent Circulation Alliance Taiwanese Talent Outbound International Internship.

Deployed genetic sequence service and developed video-based cariac organoid beating detection.

Developed a website for Taiwanese baseball data visualization, advanced analysis, and scouting purpose.

Provided data analysis service for teams in the Chinese Professional Baseball League (CPBL) and the Chinese Taipei National Baseball Team in World Baseball Classics 2023.

Website template is adapted from here.