Yunhai Tong

Papers

Total Citations

108

H-Index

About

Yunhai Tong is a leading researcher in computer vision and multimodal AI, with key contributions to glass-like object segmentation and 4D video recognition. His work on "Enhanced Boundary Learning for Glass-like Object Segmentation" (2021, 103 citations) tackles the notoriously difficult problem of detecting transparent surfaces—such as windows, bottles, and mirrors—which are critical for robot navigation and grasping. By developing novel boundary-learning techniques, Tong’s method overcomes challenges posed by arbitrary backgrounds behind glass, significantly advancing practical sensing systems. More recently, Tong introduced "VG4D: Vision-Language Model Goes 4D Video Recognition" (2024, 5 citations), a pioneering approach that integrates vision-language models with 4D point cloud video understanding. This work addresses limitations in sensor resolution for autonomous driving and robotics, enabling richer, more detailed scene interpretation by leveraging multimodal data. Tong’s research bridges the gap between 2D vision and dynamic 3D environments, with applications spanning from industrial automation to self-driving cars. His innovative boundary-learning framework has become a benchmark in transparent object segmentation, while his 4D video work opens new frontiers for real-world AI systems.