Object Detection on SUN RGB-D
36.6mAP@0.5F.RCNN (Tea.)
Evaluation results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| F.RCNN (Tea.)Backbone=Res50 (×2), Input=RGB-D2024.05 | 36.6 | 57.8 | 30.7 | 42.4 | 38.9 | 61.3 | 13.8 | — | — | — | — | |
| F.RCNN (Stu.)w/AMFDBackbone=Res18 (×1), Input=RGB-D2024.05 | 36.5 | 56 | 31.7 | 41.5 | 39.2 | 61.3 | 14.5 | — | — | — | — | |
| Contrastive Scene ContextsPre-training Initialization=Pre-trained2020.12 | 36.4 | 55.3 | 20.3 | 53.8 | 53.6 | 65.9 | 6.1 | 15.5 | 38 | 9.1 | 46.5 | |
| Contrastive Scene ContextsPre-training strategy=Contrastive Scene Contexts2020.12 | 36.4 | — | — | — | — | — | — | — | — | — | — | |
| PointContrastPre-training Initialization=Pre-trained2020.12 | 34.8 | 50.5 | 19.4 | 51.8 | 54.9 | 57.4 | 7.5 | 16.2 | 37 | 5.9 | 47.6 | |
| PointContrastPre-training strategy=PointContrast2020.12 | 34.8 | — | — | — | — | — | — | — | — | — | — | |
| F.RCNN (Stu.)Backbone=Res18 (×1), Input=RGB-D2024.05 | 32.1 | 50.9 | 28.3 | 36.5 | 34.9 | 53.3 | 11.1 | — | — | — | — | |
| ScratchPre-training Initialization=None (Random)2020.12 | 31.7 | 47.8 | 19.6 | 48.1 | 54.6 | 60 | 6.3 | 15.8 | 27.3 | 5.4 | 32.1 | |
| ScratchPre-training strategy=None (from scratch)2020.12 | 31.7 | — | — | — | — | — | — | — | — | — | — | |
| F.RCNNBackbone=Res50 (×2), Input=RGB2024.05 | 30.6 | 44.2 | 28.5 | 32.1 | 36.6 | 51.9 | 10.7 | — | — | — | — | |
| F.RCNNBackbone=Res18 (×1), Input=Depth2024.05 | 30.5 | 46.3 | 25.7 | 34.2 | 33.3 | 53.3 | 11.4 | — | — | — | — | |
| F.RCNNBackbone=Res18 (×1), Input=RGB2024.05 | 30.2 | 43.9 | 24.5 | 33.9 | 33.2 | 53.6 | 10.3 | — | — | — | — | |
| F.RCNNBackbone=Res50 (×2), Input=Depth2024.05 | 29.5 | 52.9 | 27.4 | 34.1 | 34.3 | 54.4 | 8 | — | — | — | — |