Volume Transformer: Revisiting Vanilla Transformers for 3D Scene Understanding

Benchmark Model Rank Results
3d-instance-segmentation-on-scannet200-testVolt-SPFormer#1AP50: 47.5
3d-instance-segmentation-on-scannetv2Volt-B + SPFormer#1mAP @ 50: 82.7
lidar-semantic-segmentation-on-waymo-open-datasetVolt-B (extra data)#1mIoU: 73.4
3d-semantic-segmentation-on-scannet-plus-plusVolt-B#3Top-1 IoU: 0.495
3d-semantic-segmentation-on-scannet-plus-plusVolt (Top-3)#10Top-3 IoU: 0.753
3d-semantic-segmentation-on-semantickittiVolt-B#5test mIoU: 75.2%
lidar-semantic-segmentation-on-nuscenesVolt-B#7test mIoU: 0.822
3d-semantic-segmentation-on-scannet200Volt-B#16test mIoU: 41.6
semantic-segmentation-on-scannetVolt-B#47test mIoU: 80.5