Papers
4
Total Citations
13
H-Index
2
About
Tong He is a rising researcher at the intersection of computer vision and robot learning, whose work rethinks how machines perceive and interact with the physical world. He is best known for challenging conventional wisdom in robotics through his landmark paper, "Point Cloud Matters," which systematically demonstrates that 3D point cloud observations can outperform RGB and depth modalities in robot learning tasks—a finding that has already garnered significant attention and citations. His research spans open-world video segmentation, where he developed VideoSAM to extend the Segment Anything Model’s capabilities to continuous video streams for autonomous driving and robotics, and multi-domain trajectory prediction with Tra-MoE, which learns from diverse data sources to improve robot generalization. With multiple highly-cited papers published in top venues in 2024-2025, He’s work is shaping how researchers design observation spaces and perception systems for embodied AI. His contributions are particularly notable for bridging the gap between 3D computer vision and practical robot learning, offering concrete evidence that point clouds deserve renewed attention in the era of large vision models.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3VideoSAM: Open-World Video Segmentation2 citations · 2025
- 4