Max Robotics — Global Service Robot Directory + US Certification

About

Ji Hou is a researcher whose work sits at the intersection of 3D computer vision, scene understanding, and geometric deep learning. Best known for developing **RevealNet**, a pioneering approach to inferring hidden geometry in RGB-D scans, Hou has tackled one of the most practically significant challenges in 3D reconstruction: recovering object surfaces that remain unseen during scanning. This work, which has garnered 64 citations, has direct implications for robotics, augmented reality, and autonomous navigation, where complete environmental models are critical for safe and effective operation. Hou has further advanced the field through contributions to semantic instance completion, introducing the task of jointly detecting and completing object instances from incomplete RGB-D scans, and through panoptic 3D scene reconstruction from single RGB images — a technically demanding problem combining geometry, semantics, and instance understanding in a unified framework. These efforts reflect a broader commitment to holistic, application-ready scene perception rather than isolated subtasks. More recently, Hou has explored fluid dynamics and bio-inspired locomotion through deep reinforcement learning simulations of fish schooling behavior, demonstrating an impressive breadth of interdisciplinary curiosity. Collectively, his research offers meaningful tools for the next generation of intelligent, spatially aware systems.