Papers
2
Total Citations
8
H-Index
2
About
Zhipeng Fan is a rising researcher in computer vision and robotics, whose work focuses on enabling machines to perceive, reconstruct, and interact with 3D objects in open-world environments. His key research areas include category-level 6D pose estimation, 3D shape reconstruction, and self-supervised learning for articulated objects. In his highly cited work "DTF-Net" (2023, 6 citations), Fan introduced a novel deformable template field approach that simultaneously estimates object poses and reconstructs shapes from RGB-D images, overcoming the limitations of traditional methods that fail to handle shape variations across different object instances. More recently, in "SM³" (2024, 2 citations), he pioneered a self-supervised multi-task framework that uses multi-view 2D images to reconstruct articulated objects and estimate their movable joint structures—a critical capability for robotic manipulation—without relying on expensive annotated datasets. Fan's contributions are particularly notable for moving beyond category-specific supervised learning toward more generalizable, annotation-efficient solutions. His work directly addresses fundamental challenges in open-world perception, with potential applications in autonomous robotics, augmented reality, and digital twin creation. As a young researcher, his innovative approaches to deformable templates and self-supervised learning are already shaping the future of 3D scene understanding.
Research Focus
Key Achievements
Top Papers
- 1
- 2