Papers
7
Total Citations
5,314
H-Index
7
About
Oncel Tuzel is a distinguished computer vision and machine learning researcher whose work spans 3D object detection, robotic perception, and pose estimation. He is perhaps best known as a co-author of **VoxelNet** (2018), a landmark paper that revolutionized LiDAR-based 3D object detection by introducing an end-to-end deep learning framework for processing sparse point clouds — a critical capability for autonomous vehicles and robotics. The work has garnered over 4,500 citations, cementing its status as a foundational contribution to the autonomous driving literature. Earlier in his career, Tuzel made significant strides in robotic bin-picking and pose estimation, developing practical vision systems capable of detecting and localizing objects in heavily cluttered, unstructured environments. His 2012 papers on object localization and voting-based pose estimation — each approaching 200 citations — demonstrated robust solutions for real-world industrial robotics applications, including work addressing the notoriously difficult challenge of specular or shiny objects. His creative use of multi-flash camera technology for depth-edge extraction further highlighted his innovative approach to sensing. Across both industrial robotics and modern deep learning, Tuzel's research has consistently bridged fundamental computer vision challenges with high-impact practical applications.
Research Focus
Key Achievements
Top Papers
- 1VoxelNet: End-to-End Learning for Point Cloud Based 3D Object Detection4,542 citations · 2018
- 2VoxelNet: End-to-End Learning for Point Cloud Based 3D Object Detection317 citations · 2017
- 3
- 4Voting-based pose estimation for robotic assembly using a 3D sensor192 citations · 2012
- 5Pose estimation in heavy clutter using a multi-flash camera32 citations · 2010
- 6P2Π: A Minimal Solution for Registration of 3D Points to 3D Planes19 citations · 2010
- 7Finding a needle in a specular haystack18 citations · 2011