Zero Shot Transfer 3D Point Cloud 2
评估指标
OBJ_ONLY Accuracy(%)
评测结果
各个模型在此基准测试上的表现结果
模型名称 | OBJ_ONLY Accuracy(%) | Paper Title | Repository |
|---|---|---|---|
| TAMM-PointBERT (+dlign) | 60.5 | OpenDlign: Open-World Point Cloud Understanding with Depth-Aligned Images | - |
| PointCLIP V2 | 50.09 | PointCLIP V2: Prompting CLIP and GPT for Powerful 3D Open-world Learning | - |
| ReCon++ | 65.4 | ShapeLLM: Universal 3D Object Understanding for Embodied Interaction | - |
| ReCon | 43.7 | Contrast with Reconstruct: Contrastive 3D Representation Learning Guided by Generative Pretraining | - |
| OpenDlign | 59.5 | OpenDlign: Open-World Point Cloud Understanding with Depth-Aligned Images | - |
| ViT-Lens | 60.1 | ViT-Lens: Initiating Omni-Modal Exploration through 3D Insights | - |
| CLIP2Point | 30.46 | CLIP2Point: Transfer CLIP to Point Cloud Classification with Image-Depth Pre-training | - |
| Uni3D | 65.3 | Uni3D: Exploring Unified 3D Representation at Scale | - |
| PointCLIP | 19.28 | PointCLIP: Point Cloud Understanding by CLIP | - |
| MixCon3D-PointBERT | 58.6 | Sculpting Holistic 3D Representation in Contrastive Language-Image-3D Pre-training | - |
0 of 10 row(s) selected.