Back to the Feature: Classical 3D Features are (Almost) All You Need for 3D Anomaly Detection

Benchmark Model Rank Results
3d-anomaly-detection-and-segmentation-on-mvtecBack to the Feature: Classical 3D Features are (Almost) All You Need for 3D Anomaly Detection (FPFH)#7Detection AUROC: 0.782Segmentation AUROC: 0.978
3d-anomaly-detection-on-anomaly-shapenet10BTF (FPFH)#3O-AUROC: 0.632P-AUROC: 0.790
3d-anomaly-detection-on-anomaly-shapenet10BTF (Raw)#7O-AUROC: 0.500P-AUROC: 0.515
3d-anomaly-detection-on-real-3d-adBTF (Raw)#8Mean Performance of P. and O.: 0.6785Point AUROC: 0.722
3d-anomaly-detection-on-real-3d-adBTF (FPFH)#14Mean Performance of P. and O.: 0.5845Point AUROC: 0.566
depth-anomaly-detection-and-segmentation-onBack to the Feature: Classical 3D Features are (Almost) All You Need for 3D Anomaly Detection (SIFT)#5Segmentation AUPRO: 0.910Detection AUROC: 0.727
depth-anomaly-detection-and-segmentation-onBack to the Feature: Classical 3D Features are (Almost) All You Need for 3D Anomaly Detection (HoG)#7Segmentation AUPRO: 0.771Detection AUROC: 0.559
depth-anomaly-detection-and-segmentation-onBack to the Feature: Classical 3D Features are (Almost) All You Need for 3D Anomaly Detection (Depth iNet)#8Segmentation AUPRO: 0.755Detection AUROC: 0.675
depth-anomaly-detection-and-segmentation-onBack to the Feature: Classical 3D Features are (Almost) All You Need for 3D Anomaly Detection (NSA)#9Segmentation AUPRO: 0.5572Detection AUROC: 0.696
depth-anomaly-detection-and-segmentation-onBack to the Feature: Classical 3D Features are (Almost) All You Need for 3D Anomaly Detection (RaW)#10Segmentation AUPRO: 0.442Detection AUROC: 0.573
rgb-3d-anomaly-detection-and-segmentation-onBack to the Feature: Classical 3D Features are (Almost) All You Need for 3D Anomaly Detection (BTF)#6Detection AUCROC: 0.865Segmentation AUPRO: 0.959