PointPillars
A point-cloud-based 3D object detection network that divides the point cloud into vertical pillars, extracts pillar features with PointNet, and then generates detection boxes through a 2D convolutional network; it has a simple structure and fast inference speed and is often used as a LiDAR detection baseline model.
Timeline (last 90 days)
The study uses a PointPillars ensemble model and conducts experiments on two public datasets, KITTI and nuScenes.
PointPillars is a point-cloud-based 3D object detection network that partitions the point cloud into vertical pillars, extracts pillar features with PointNet, and then generates detection boxes through a 2D convolutional network.
The experimental evaluation covers multiple triggering-condition transfer scenarios, including long-range transfer, rainy-condition transfer, and transfer between two independently trained detectors.
torch-pointcloud supports detection task models, including VoteNet, 3DETR, PointPillars, SECOND, PointRCNN, SPVCNN, and OctFormer.
All Facts (4)
The study uses a PointPillars ensemble model and conducts experiments on two public datasets, KITTI and nuScenes.
50%UnverifiedPointPillars is a point-cloud-based 3D object detection network that partitions the point cloud into vertical pillars, extracts pillar features with PointNet, and then generates detection boxes through a 2D convolutional network.
50%UnverifiedThe experimental evaluation covers multiple triggering-condition transfer scenarios, including long-range transfer, rainy-condition transfer, and transfer between two independently trained detectors.
50%Unverifiedtorch-pointcloud supports detection task models, including VoteNet, 3DETR, PointPillars, SECOND, PointRCNN, SPVCNN, and OctFormer.
50%