Learning-Deep-Learning

2D Car Detection in Radar Data with PointNets

July 2019

tl;dr: Use F-pointnet for car detection with sparse 4D radar data (x, y, $\tilde {v}_r$, $\sigma$).

Overall impression

From U of Ulm. Only one target per per car, in a controlled environment. A high precision GPS is used to create the dataset GT.

This is an extension to the radar point cloud segmentation.

Key ideas

Technical details