LCPFormer: Towards Effective 3D Point Cloud Analysis via Local Context Propagation in Transformers

Benchmark Model Rank Results
3d-semantic-segmentation-on-sensaturbanLCPFormer#1mIoU: 63.4
3d-object-detection-on-sun-rgbd-valLCPFormer#16mAP@0.25: 63.2mAP@0.5: 46.2
semantic-segmentation-on-s3dis-area5LCPFormer#32mIoU: 70.2mAcc: 76.8oAcc: 90.8Number of params: N/A
3d-point-cloud-classification-on-modelnet40LCPFormer#54Overall Accuracy: 93.6Mean Accuracy: 90.7