Yi Hong

Papers

Total Citations

H-Index

About

Yi Hong is a leading researcher in robot learning and human-robot interaction, with a focus on enabling machines to acquire skills by observing human demonstrations. Her core research areas include learning from observation (LfO), imitation learning, and state-action recognition from multimodal sensory data. Hong’s major contribution is the development of SA-Net, a deep neural network architecture that robustly recognizes state-action pairs from RGB-D video streams—a critical step for robots to decompose complex human tasks into learnable sequences. Her 2020 paper on SA-Net, which has garnered 24 citations, demonstrates how robots can autonomously segment and interpret streaming sensory data without explicit programming, advancing the paradigm of learning from observation. In her 2019 work, she further extended this approach to trajectory recognition, achieving 6 citations. Hong’s research bridges computer vision and robotics, offering practical solutions for real-world skill transfer. Her work is particularly notable for its potential to make robot learning more accessible, reducing the need for expert programming and enabling non-specialists to teach robots through natural demonstration.