Shuaicheng Liu

University of Electronic Science and Technology of China, Vi Technology (United States)

Papers

Total Citations

158

H-Index

About

Shuaicheng Liu is a versatile computer vision researcher whose work spans video processing, computational photography, and robot learning. He is perhaps best known for his pioneering contribution to video stabilization, with his 2016 paper "MeshFlow: Minimum Latency Online Video Stabilization" accumulating 149 citations and establishing him as a leading voice in real-time video processing. MeshFlow introduced a spatially smooth mesh-based motion representation that significantly reduced latency while improving stabilization quality — a breakthrough with lasting practical impact for mobile and consumer video applications. More recently, Liu has expanded his research horizons into robotics and embodied AI. His 2025 work, "FlowPolicy," addresses a critical challenge in vision-based imitation learning: enabling robots to acquire complex manipulation skills both quickly and robustly. By leveraging consistency flow matching, FlowPolicy advances the frontier of diffusion- and flow-based policy generation for robotic manipulation, demonstrating Liu's ability to translate cutting-edge generative modeling techniques into real-world robotic systems. Across his career, Liu exemplifies a researcher who bridges foundational computer vision theory with high-impact applications — from stabilizing everyday videos to teaching robots dexterous, human-like skills.