Hwann-Tzong Chen

Papers

Total Citations

H-Index

About

Hwann-Tzong Chen is a leading researcher in computer vision and machine learning, with a focus on generative models, scene understanding, and robotic perception. His work spans video generation, indoor layout estimation, and amodal segmentation, pushing the boundaries of visual AI. Chen’s 2019 paper “Point-to-Point Video Generation” (22 citations) tackles the challenging task of temporally coherent video synthesis, advancing beyond static image generation to enable real-world applications like video editing. In the same year, his “Flat2Layout” (16 citations) introduced a novel flat representation for estimating room layouts from single images, overcoming the limitations of box-shaped room assumptions and enabling general indoor scene understanding. Chen also contributed to robotics with “Chess Recognition from a Single Depth Image” (9 citations), integrating learning-based depth perception into a dual-arm robotic chess system. His latest work, “Segment Anything, Even Occluded” (2025), addresses amodal instance segmentation, detecting both visible and occluded object parts—critical for autonomous driving and robotic manipulation. With a portfolio that bridges foundational research and practical applications, Chen’s work has garnered over 49 citations, demonstrating his impact on both academic and applied computer vision.