Daniel Yang

University of Southern California, Samsung (United States)

Papers

Total Citations

144

H-Index

About

Daniel Yang is a leading researcher in robotic manipulation and computer vision, with a focus on enabling robots to understand and interact with objects in human-like ways. His work bridges the gap between perception and action, particularly through learning from visual demonstrations. Yang’s highly cited 2018 paper, “Demo2Vec: Reasoning Object Affordances from Online Videos” (108 citations), introduced a novel framework that extracts feature embeddings from demonstration videos to infer the functional affordances of unseen objects—a critical step for generalizable robotic reasoning. He further advanced the field with his 2021 work on “Robotic Grasping through Combined Image-Based Grasp Proposal and 3D Reconstruction” (36 citations), which integrates a learned grasp proposal network with a 3D shape reconstruction network to generate precise 6-DOF grasps from a single RGB-D image. This approach leverages geometric understanding to improve grasp success on novel objects. Yang’s contributions have significant implications for autonomous systems, from household robots to industrial automation, and his work is widely cited for its innovative fusion of learning-based vision and robotics. He is recognized for pushing the boundaries of affordance reasoning and data-efficient manipulation.