Beyond local patches: Preserving global–local interactions by enhancing self-attention via 3D point cloud tokenization

Open paper
Benchmark Model Rank Results
3d-part-segmentation-on-shapenet-partOurs–Instance Average IoU: 88.1
3d-point-cloud-classification-on-scanobjectnnOurs–Overall Accuracy: 89.0Mean Accuracy: 87.2
semantic-segmentation-on-s3dis-area5Ours–mIoU: 73.6mAcc: 80.2oAcc: 93.0

Rank counts only results with a code link.