ULIP: Learning a Unified Representation of Language, Images, and Point Clouds for 3D Understanding

Open paper
Benchmark Model Rank Results
training-free-3d-point-cloud-classification-on-modelnet40ULIP#4Accuracy (%): 60.4Need 3D Data?: Yes
3d-point-cloud-classification-on-modelnet40ULIP + PointMLP#7Overall Accuracy: 94.7Mean Accuracy: 92.4
3d-point-cloud-classification-on-modelnet40ULIP + PointBERT#23Overall Accuracy: 94.1
3d-point-cloud-classification-on-modelnet40ULIP + PointNet++(ssg)#63Overall Accuracy: 93.4Mean Accuracy: 91.2
zero-shot-transfer-3d-point-cloud-classification-on-modelnet40ULIP + PointMLP#13Accuracy (%): 61.5
zero-shot-transfer-3d-point-cloud-classification-on-modelnet40ULIP + PointBERT#14Accuracy (%): 60.4
3d-point-cloud-classification-on-scanobjectnnULIP + PointNeXt#22Overall Accuracy: 89.7Mean Accuracy: 88.6Number of params: 1.4M
3d-point-cloud-classification-on-scanobjectnnULIP + PointMLP#25Overall Accuracy: 89.4Mean Accuracy: 88.5
3d-point-cloud-classification-on-scanobjectnnULIP + PointBERT#48Overall Accuracy: 86.4

Rank counts only results with a code link.