Hehe Fan

University of Technology Sydney, National University of Singapore, Zhejiang University

Papers

Total Citations

H-Index

About

Hehe Fan is a researcher pushing the boundaries of how machines perceive, predict, and interact with the physical world. His work spans three critical frontiers: video prediction, cross-modal localization, and 3D hand-object interaction. Fan’s most influential contribution is **“Cubic LSTMs for Video Prediction”** (2019, 46 citations), which introduced a novel architecture to capture complex spatiotemporal dynamics for anticipating future video frames—a cornerstone for robotics and autonomous systems. He also pioneered **“Text to Point Cloud Localization with Relation-Enhanced Transformer”** (2023, 11 citations), enabling robots to pinpoint locations from natural language descriptions, bridging human communication and 3D spatial understanding. In **“Hand-Centric Motion Refinement for 3D Hand-Object Interaction”** (2024, 5 citations), Fan addressed the challenge of generating realistic hand motion during object manipulation, vital for VR and robotic dexterity. His latest work, **“TSGS: Improving Gaussian Splatting for Transparent Surface Reconstruction”** (2025, 3 citations), tackles the notoriously difficult problem of reconstructing transparent objects, directly impacting lab robotics and scene understanding. With over 65 citations across his key publications, Fan is recognized for developing practical, high-impact solutions that advance embodied AI and human-robot collaboration.