Towards Large-scale 3D Representation Learning with Multi-dataset Point Prompt Training

Benchmark Model Rank Results
semantic-segmentation-on-s3disPPT + SparseUNet#7Mean IoU: 78.1mAcc: 85.4oAcc: 92.2Number of params: N/A
semantic-segmentation-on-scannetPPT + SparseUNet#10val mIoU: 76.4test mIoU: 76.6
3d-semantic-segmentation-on-scannet200PPT+SparseUNet#11val mIoU: 31.9test mIoU: 33.2
semantic-segmentation-on-s3dis-area5PPT + SparseUNet#14mIoU: 72.7mAcc: 78.2oAcc: 91.5Number of params: N/A
lidar-semantic-segmentation-on-nuscenesPPT+SparseUNet#22val mIoU: 0.786
3d-semantic-segmentation-on-semantickittiPPT+SparseUNet#40val mIoU: 71.4%