Results for TransformerRefer
Submission data
| Full name | ScanRefer with Transformer-based Object Detection |
| Description | ScanRefer method, but replaced VoteNet with Group-Free Transformer-based 3D Object Detector.
Uses their provided weights for the 12 layer, double width, 256 candidates model as initialization, and therefor only XYZ (without height) as input features. |
| Input Data Types | Uses XYZ coordinates |
| Programming language(s) | Python |
| Hardware | GeForce RTX 2080 Ti, 11GB RAM |
| Submission creation date | 8 Jul, 2021 |
| Last edited | 8 Jul, 2021 |
Localization
| Unique | Unique | Multiple | Multiple | Overall | Overall |
|---|---|---|---|---|---|
| acc@0.25IoU | acc@0.5IoU | acc@0.25IoU | acc@0.5IoU | acc@0.25IoU | acc@0.5IoU |
| 0.6010 | 0.4658 | 0.2540 | 0.1730 | 0.3318 | 0.2386 |