Object Recognition on Objaverse-LVIS
53.7Top-1 AccRECON++-L
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| RECON++-LPre-training Data=Ensembled, Scale=Large2024.02 | 53.7 | 75.8 | 82 | — | |
| RECON++-BPre-training Data=Ensembled, Scale=Base2024.02 | 53.2 | 75.3 | 81.5 | — | |
| Uni3D-LPre-training Data=Ensembled, Teacher=EVA-CLIP-E2024.02 | 53.1 | 75 | 81.5 | — | |
| MixCon3DPre-training Data=Ensembled2024.02 | 52.5 | 74.5 | 81.2 | — | |
| Uni3D-BPre-training Data=Ensembled, Teacher=EVA-CLIP-E2024.02 | 51.7 | 74.1 | 80.8 | — | |
| TAMMPre-training Data=Ensembled2024.02 | 50.7 | 73.2 | 80.6 | — | |
| SUGAR (single)Zero-shot=true, Training Data=Ensembled w/ LVIS dataset2024.04 | 49.5 | — | 78.8 | — | |
| OpenShapePre-training Data=Ensembled2024.02 | 46.8 | 69.1 | 77 | — | |
| OpenShapeZero-shot=true, Training Data=Ensembled w/ LVIS dataset2024.04 | 46.8 | — | 77 | — | |
| SUGAR (multi)Zero-shot=true, Training Data=Ensembled w/ LVIS dataset2024.04 | 46.8 | — | 76.6 | — | |
| TIGAUSSIAN3D Representation=3DGS, Evaluation Protocol=Zero-shot2026.01 | 41.76 | 62.68 | 69.15 | — | |
| Duoduo CLIP3D Representation=Multi-view (non explicit), Evaluation Protocol=Zero-shot2026.01 | 38.05 | 57.79 | 66.7 | — | |
| UniGS3D Representation=3DGS, Evaluation Protocol=Zero-shot2026.01 | 37.64 | 57.62 | 65.57 | — | |
| Uni3D3D Representation=Point Cloud, Evaluation Protocol=Zero-shot2026.01 | 36.72 | 57.09 | 65.18 | — | |
| ULIP-23D Representation=Point Cloud, Evaluation Protocol=Zero-shot2026.01 | 29.75 | 46.39 | 55.23 | — | |
| ULIP-2Pre-training Data=Ensembled2024.02 | 26.8 | 44.8 | 52.6 | — | |
| MixCon3DPre-training Data=ShapeNet2024.02 | 22.3 | 37.5 | 44.3 | — | |
| TAMMPre-training Data=ShapeNet2024.02 | 13.7 | 24.2 | 29.2 | — | |
| CLIP3D Representation=Point Cloud, Evaluation Protocol=Zero-shot2026.01 | 12.35 | 24.62 | 32.91 | — | |
| OpenShapePre-training Data=ShapeNet2024.02 | 10.8 | 20.2 | 25 | — | |
| ULIPPre-training Data=ShapeNet2024.02 | 6.2 | 13.6 | 17.9 | — | |
| PointCLIPv2Protocol=2D Inference without 3D Training2024.02 | 4.7 | 9.5 | 12.9 | — | |
| CLIP2PointPre-training Data=ShapeNet2024.02 | 2.7 | 5.8 | 7.9 | — | |
| PointCLIPProtocol=2D Inference without 3D Training2024.02 | 1.9 | 4.1 | 5.8 | — | |
| RECONPre-training Data=ShapeNet2024.02 | 1.1 | 2.7 | 3.7 | — |