Haoda Li

University of Toronto

Papers

2

Total Citations

38

H-Index

2

About

Haoda Li is a researcher at the forefront of geometric deep learning and robotic manipulation, with a primary focus on enabling machines to autonomously understand and assemble 3D shapes. His most notable contribution is the pioneering work "Neural Shape Mating: Self-Supervised Object Assembly with Adversarial Shape Priors" (2022), which has garnered significant attention with 34 citations. This research redefines the challenge of object assembly by moving beyond merely posing semantic parts to recreating a whole; instead, Li introduces a self-supervised framework that learns to "mate" geometric parts together based purely on their shape compatibility. By employing adversarial shape priors, his method allows robots to autonomously fit broken or separated pieces together without requiring labeled data, a critical step toward practical robotic repair and assembly in unstructured environments. Li's work has profound implications for manufacturing, automated disassembly, and even archaeological reconstruction, demonstrating how deep learning can bridge the gap between perception and physical interaction. His innovative approach to self-supervised learning in 3D geometry positions him as a rising leader in the intersection of computer vision and robotics.

Research Focus

Key Achievements

2
H-Index
2
Papers
38
Total Citations
19
Avg Citations/Paper
🏆 Most Cited Paper
Neural Shape Mating: Self-Supervised Object Assembly with Adversarial Shape Priors
34 citations · 2022
📈 Most Prolific Year: 2022 (2 Papers)
🤝 Key Collaborators: 4
🏛 Institutions: University of Toronto

Top Papers

  1. 1
  2. 2

Key Collaborators

Contact & Links

Available for collaboration
Content generated · 13 days ago