Jia Deng

University of Michigan–Ann Arbor, Ann Arbor Center for Independent Living

Papers

Total Citations

H-Index

About

Jia Deng’s research lies at the intersection of computer vision, robotics, and human-computer interaction, with a focus on enabling machines to understand the physical world through visual data. His work on mining semantic affordances of visual object categories—awarded 65 citations—has been foundational in defining how objects communicate their functionality, such as the graspability of a cup or the sittability of a chair. This insight is critical for robots that must interact with objects and for systems that recognize human activities. Deng also advanced spatial reasoning with the Rel3D benchmark (13 citations), which provides high-quality 3D ground truth data for grounding spatial relations like “on” or “under,” addressing a key gap in existing datasets. In a notable interdisciplinary achievement, he contributed to the VAST project (6 citations), applying machine learning to video analysis of surgeons performing robotic prostatectomy, demonstrating how computer vision can assess and improve surgical skill. Through these contributions, Deng has shaped how machines perceive object functionality, spatial context, and even human expertise, making his work impactful across robotics, AI, and medical training.