Point-Voxel CNN for Efficient 3D Deep Learning

Benchmark Model Rank Results
3d-object-detection-on-kitti-pedestrian-easy-valPVCNN#1AP: 73.2
3d-object-detection-on-kitti-pedestrian-hard-valPVCNN#1AP: 56.78
3d-object-detection-on-kitti-pedestrian-moderate-valPVCNN#1AP: 64.71
3d-object-detection-on-kitti-cyclist-easy-valPVCNN#2AP: 81.4
3d-object-detection-on-kitti-cyclist-hard-valPVCNN#2AP: 56.24
3d-object-detection-on-kitti-cyclist-moderate-valPVCNN#2AP: 59.97
3d-object-detection-on-kitti-cars-easy-valPVCNN#5AP: 84.02
3d-object-detection-on-kitti-cars-hard-valPVCNN#5AP: 63.81
3d-object-detection-on-kitti-cars-moderate-valPVCNN#6AP: 71.54
3d-semantic-segmentation-on-s3disPVCNN++#6mIoU (6-Fold): 58.98
3d-part-segmentation-on-shapenet-partPVCNN volumetric#30Instance Average IoU: 86.2