3D Object Recognition on ShapeNet
99.59AccuracyOurs w/ PointBind
Evaluation Results
| Method | Links | |
|---|---|---|
| Ours w/ PointBindSetting=Fine-tuning2024.03 | 99.59 | |
| Meta-TransformerSetting=Fine-tuning2024.03 | 99.3 | |
| PointBindSetting=Fine-tuning, Protocol=linear2024.03 | 99.09 | |
| PointBind (+Event)Setting=Fine-tuning2024.03 | 99.09 | |
| Ours w/ PointBindSetting=Zero-shot2024.03 | 98.96 | |
| PointBindSetting=Zero-shot2024.03 | 98.85 | |
| PointBind (+Event)Setting=Zero-shot2024.03 | 98.85 | |
| LLaVA-vicuna-13bModality=Image (Multi-view)2024.06 | 73.45 | |
| LLaNA-7bModality=NeRF2024.06 | 67.14 | |
| LLaVA-vicuna-13bModality=Image (Front-view)2024.06 | 66.13 | |
| LLaVA-vicuna-13bModality=Image (Back-view)2024.06 | 63.9 | |
| BLIP-2 FlanT5-xxlModality=Image (Front-view)2024.06 | 63.67 | |
| BLIP-2 FlanT5-xxlModality=Image (Back-view)2024.06 | 61.47 | |
| MetaGPBackbone=DGCNN, # Classes=16, Reusing Method=MetaGP2023.04 | 60.9 | |
| 3D-LLMModality=Mesh + Multi-view2024.06 | 60.55 | |
| LLaVA-vicuna-7bModality=Image (Front-view)2024.06 | 60.25 | |
| LLaVA-vicuna-7bModality=Image (Back-view)2024.06 | 57 | |
| PointLLM-7bModality=Point cloud2024.06 | 50.14 | |
| GPT4Point-Opt-2.7bModality=Point cloud2024.06 | 41.93 | |
| Vanilla ReusingBackbone=DGCNN, # Classes=16, Reusing Method=Vanilla2023.04 | 15.45 |