Surface Representation for Point Clouds

Benchmark Model Rank Results
3d-object-detection-on-scannetv2RepSurf-U#14mAP@0.5: 54.8mAP@0.25: 71.2
3d-object-detection-on-sun-rgbd-valRepSurf-U#11mAP@0.25: 64.9mAP@0.5: 47.7
3d-point-cloud-classification-on-modelnet40RepSurf-U#5Overall Accuracy: 94.7Number of params: 1.48MFLOPs: 0.81G
3d-point-cloud-classification-on-scanobjectnnRepSurf-U (2x)#47Overall Accuracy: 86.0FLOPs: 2.43GNumber of params: 6.80M
3d-point-cloud-classification-on-scanobjectnnRepSurf-U#53Overall Accuracy: 84.6FLOPs: 0.81GNumber of params: 1.48M
semantic-segmentation-on-s3disRepSurf-U#14Mean IoU: 74.3mAcc: 82.6oAcc: 90.8FLOPs: 1.04G
semantic-segmentation-on-s3dis-area5RepSurf-U#28mIoU: 68.9mAcc: 76.0oAcc: 90.2FLOPs: 1.04G