3D Captioning on Objaverse (test)
100S-BERT ScoreHuman
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| HumanSource=Human Annotation2023.08 | 100 | 100 | 100 | 2.67 | 0.22 | 92.46 | 100 | 100 | 100 | |
| PnP-U3DInput modality=2D+3D2026.02 | 61.03 | — | 64.8 | — | — | — | 19.75 | 19.97 | 16.41 | |
| PointLLMModel scale=7B2026.02 | 55.76 | — | 58.27 | — | — | — | 8.7 | 11.07 | 13.25 | |
| PointLLMModel scale=13B2026.02 | 55.5 | — | 57.96 | — | — | — | 8.53 | 11 | 13.08 | |
| PointLLM-13BModel Scale=13B, Caption length constraint=no more than 20 words2023.08 | 50.15 | 44.27 | 50.83 | 2.12 | 0.39 | 84.39 | 17.09 | 20.99 | 16.45 | |
| SHAPELLM-13BInput=3D Point Cloud2024.02 | 48.52 | 48.94 | 49.98 | — | — | — | — | — | — | |
| SHAPELLM-7BInput=3D Point Cloud2024.02 | 48.2 | 46.92 | 49.23 | — | — | — | — | — | — | |
| PointLLM-13BInput=3D Point Cloud2024.02 | 47.91 | 48.15 | 49.12 | — | — | — | — | — | — | |
| PointLLM-13BModel Scale=13B2023.08 | 47.91 | 48.15 | 49.12 | 3.1 | 0.84 | 78.75 | 3.83 | 7.23 | 12.26 | |
| PointLLM-7BInput=3D Point Cloud2024.02 | 47.47 | 44.85 | 48.55 | — | — | — | — | — | — | |
| PointLLM-7BModel Scale=7B2023.08 | 47.47 | 44.85 | 48.55 | 3.04 | 0.66 | 82.14 | 3.87 | 7.3 | 11.92 | |
| InstructBLIP-7BInput=1-View 2D Image2024.02 | 47.41 | 45.34 | 48.48 | — | — | — | — | — | — | |
| InstructBLIP-7BModel Scale=7B2023.08 | 47.41 | 45.34 | 48.48 | 2.56 | 0.77 | 76.99 | 4.27 | 8.28 | 12.99 | |
| LLaVA-13BInput=1-View 2D Image2024.02 | 46.37 | 38.28 | 45.9 | — | — | — | — | — | — | |
| LLaVA-13BModel Scale=13B2023.08 | 46.37 | 38.28 | 45.9 | 2.43 | 0.86 | 73.97 | 4.02 | 8.15 | 12.58 | |
| InstructBLIP-13BInput=1-View 2D Image2024.02 | 45.9 | 44.97 | 48.86 | — | — | — | — | — | — | |
| InstructBLIP-13BModel Scale=13B2023.08 | 45.9 | 44.97 | 48.86 | 2.58 | 1.13 | 69.56 | 4.65 | 8.85 | 13.23 | |
| PnP-U3DInput modality=3D2026.02 | 45.8 | — | 47.42 | — | — | — | 12.86 | 15.38 | 10.26 | |
| LLaVA-7BInput=1-View 2D Image2024.02 | 45.61 | 46.71 | 47.1 | — | — | — | — | — | — | |
| LLaVA-7BModel Scale=7B2023.08 | 45.61 | 46.71 | 47.1 | 2.76 | 0.86 | 76.3 | 3.64 | 7.7 | 12.14 | |
| 3D-LLMInput=3D Object + Multi-View 2D Image2024.02 | 44.48 | 33.42 | 43.68 | — | — | — | — | — | — | |
| 3D-LLMModel Scale=not specified2023.08 | 44.48 | 33.42 | 43.68 | 1.77 | 1.16 | 60.39 | 16.91 | 19.48 | 19.73 | |
| QwenVL-2.5Input modality=2D2026.02 | 42.75 | — | 47.93 | — | — | — | 8.04 | 8.33 | 6.27 | |
| ShapeLLM-OmniInput modality=Omni2026.02 | 35.71 | — | 38.66 | — | — | — | 13.55 | 16.44 | 10.5 | |
| PointLLM-7bModality=Point cloud2024.06 | 33.88 | — | 33.04 | — | — | — | 5.37 | 8.14 | 12.28 | |
| LLaNA-7bModality=NeRF2024.06 | 30.07 | — | 28.17 | — | — | — | 4.93 | 7.47 | 11.15 | |
| LLaVA-vicuna-13bModality=Image (FV)2024.06 | 27.07 | — | 26.82 | — | — | — | 4.41 | 6.81 | 9.77 | |
| GPT4Point-Opt-2.7bModality=Point cloud2024.06 | 25.94 | — | 29.04 | — | — | — | 4.25 | 7.99 | 10.42 |