Contrast with Reconstruct: Contrastive 3D Representation Learning Guided by Generative Pretraining

Benchmark Model Rank Results
zero-shot-transfer-3d-point-cloud-classification-on-modelnet10ReCon#1Accuracy (%): 75.6
3d-point-cloud-linear-classification-on-modelnet40ReCon#4Overall Accuracy: 93.4
3d-point-cloud-classification-on-modelnet40ReCon#8Overall Accuracy: 94.7
few-shot-3d-point-cloud-classification-on-modelnet40-10-way-20-shotReCon#8Overall Accuracy: 95.8Standard Deviation: 3.0
few-shot-3d-point-cloud-classification-on-modelnet40-5-way-20-shotReCon#8Overall Accuracy: 98.9Standard Deviation: 1.2
zero-shot-transfer-3d-point-cloud-classification-on-scanobjectnnReCon#8OBJ_ONLY Accuracy(%): 43.7OBJ_BG Accuracy(%): 40.4…
few-shot-3d-point-cloud-classification-on-modelnet40-5-way-10-shotReCon#9Overall Accuracy: 97.3Standard Deviation: 1.9
3d-point-cloud-classification-on-scanobjectnnReCon#10Overall Accuracy: 91.26OBJ-BG (OA): 95.35OBJ-ONLY (OA): 93.80
3d-point-cloud-classification-on-scanobjectnnReCon (no voting)#12Overall Accuracy: 90.63OBJ-BG (OA): 95.18OBJ-ONLY (OA): 93.29
zero-shot-transfer-3d-point-cloud-classification-on-modelnet40ReCon#12Accuracy (%): 61.7
few-shot-3d-point-cloud-classification-on-modelnet40-10-way-10-shotReCon#13Overall Accuracy: 93.3Standard Deviation: 3.9