3D Point-to-Text Retrieval on ObjaverseXL-LVIS 1K (test)
32.2R@1GPT4Point
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GPT4PointInput Data Type=Point Cloud (+Color)2023.12 | 32.2 | 64 | 81.3 | |
| InstructBLIPInput Data Type=Mesh with Color2023.12 | 20.4 | 43.1 | 55.3 | |
| BLIP-2Input Data Type=Single-View Image2023.12 | 17.56 | 41.16 | 52.82 |