Visual data in autonomous driving perception, such as camera image and LiDAR point cloud, can be interpreted as a mixture of two aspects: semantic feature and geometric structure. Semantics come from the appearance and context of objects to the sensor, while geometric structure is the actual 3D shape of point clouds. Most detectors on LiDAR point clouds focus only on analyzing the geometric structure of objects in real 3D space. Unlike previous works, we propose to learn both semantic feature and geometric structure via a unified multi-view framework. Our method exploits the nature of LiDAR scans -- 2D range images, and applies well-studied 2D convolutions to extract semantic features. By fusing semantic and geometric features, our method outperforms state-of-the-art approaches in all categories by a large margin. The methodology of combining semantic and geometric features provides a unique perspective of looking at the problems in real-world 3D point cloud detection.

PanoNet3D's framework for point cloud detection]{PanoNet3D's framework for point cloud detection. The top branch takes LiDAR point cloud as input and decorates raw point features with several simple local geometric features. The lower branch converts point cloud to pseudo range image and feeds it into a 2D FCN to get per-pixel deep semantic feature. The output features of these two branches are then aggregated and passed to the main detector. A final bounding box head generates detected proposals on the BEV plane.

@inproceedings{xia20panonet3d,
    Author = {Chen, Xia and Wang, Jianren and Held, David and Hebert, Martial},
    Title = {PanoNet3D: Combining Semantic and Geometric Understanding
    for LiDARPoint Cloud Detection},
    Booktitle = {3DV},
    Year = {2020}
}