Congratulations Dr. Jialong Wu
Yesterday was a big day for our OSVIA Lab at Universität Osnabrück: Jialong Wu successfully defended his PhD thesis “Radar-Centric Perception for Autonomous Driving: Object Detection, Scene Flow, and Radar-Camera Fusion”, becoming the first PhD graduate of our lab, just one year after it was founded.
Jialong’s industrial PhD, carried out in cooperation with Aptiv, resulted in three original contributions to learning-based radar perception:
1) SparseRadNet (ECCV 2024): subsamples raw radar data down to about 3% of its pixels and still reaches state of the art object detection performance.
https://osvia.org/Sparse-RadNet/
2) TARS (ICCV 2025): estimates 3D scene flow from sparse radar point clouds by combining point level motion cues with a learned representation of traffic context.
3) ConFusion (currently under review): a new radar-camera fusion paradigm for 3D object detection built around explicit interaction between heterogeneous image, radar, and world queries.
https://arxiv.org/abs/2604.25574
Congratulations to Jialong Wu on an excellent thesis and defense. Thanks also to the examination committee, Holger Caesar (TU Delft), Joachim Hertzberg (Osnabrück University & DFKI), and Annika Mütze from my own group, for a great discussion.