Cylindrical and Asymmetrical 3D Convolution Networks for LiDAR Segmentation

Benchmark Model Rank Results
3d-semantic-segmentation-on-scribblekittiCylinder3D#2mIoU: 57.0
3d-semantic-segmentation-on-semantickittiCylinder3D#10test mIoU: 68.9%val mIoU: 64.3%
lidar-semantic-segmentation-on-nuscenesCylinder3D+InstanceAug#11test mIoU: 0.77
lidar-semantic-segmentation-on-s-midCylinder3D#2val mIoU: 68.8%
robust-3d-semantic-segmentation-on-nuscenes-cCylinder3D (torchsparse)#6mean Corruption Error (mCE): 105.56%
robust-3d-semantic-segmentation-on-nuscenes-cCylinder3D (spconv)#9mean Corruption Error (mCE): 111.84%
robust-3d-semantic-segmentation-on-robo3dCylinder3D (torchsparse)#7mean Corruption Error (mCE): 103.13%
robust-3d-semantic-segmentation-on-robo3dCylinder3D (spconv)#8mean Corruption Error (mCE): 103.25%
robust-3d-semantic-segmentation-on-wod-cCylinder3D (torchsparse)#5mean Corruption Error (mCE): 106.02%
semi-supervised-semantic-segmentation-on-23Sup.-only (Voxel)#3mIoU (1% Labels): 39.2mIoU (10% Labels): 48.0mIoU (20% Labels): 52.1
semi-supervised-semantic-segmentation-on-25Sup.-only (Voxel)#5mIoU (1% Labels): 50.9mIoU (10% Labels): 65.9mIoU (20% Labels): 66.6