3D Text-to-Point Retrieval on ObjaverseXL-LVIS 1K (test)
89.7Recall@1GPT4Point
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GPT4PointInput Data Type=Point Cloud (+Color)2023.12 | 89.7 | 98.1 | 98.9 | |
| BLIP-2Input Data Type=Single-View Image2023.12 | 16.72 | 40.2 | 52.56 | |
| InstructBLIPInput Data Type=Mesh with Color2023.12 | 13.7 | 32.5 | 42.7 |