Papers
2
Total Citations
7
H-Index
1
About
Eddy Ilg is a computer vision researcher whose work bridges 3D perception, motion estimation, and robotics. His research focuses on reconstructing dynamic scenes from sensor data, with key contributions in optical flow for 3D point cloud acquisition and category-level object pose estimation. Ilg’s early work tackled the challenge of reconstructing rigid body models from laser range data distorted by motion, using optical flow to correct for moving objects during 3D scanning—a problem critical for autonomous systems operating in dynamic environments. More recently, he has advanced unsupervised learning for category-level 3D pose estimation from object-centric videos, enabling robots and embodied agents to understand object orientation without human annotations or CAD models. While his most-cited paper has 6 citations, his work addresses fundamental bottlenecks in 3D vision, particularly the need for scalable, annotation-free methods. Ilg’s contributions are notable for pushing toward practical, real-world deployment of perception systems, where robustness to motion and minimal supervision are essential. His research continues to impact fields like robotics, autonomous navigation, and 3D generative modeling.
Research Focus
Key Achievements
Top Papers
- 1
- 2