2D Semantic Segmentation on ScanNet (val)
66.7mIoUMask3D (MAE)
Evaluation results
| Method | Links | |
|---|---|---|
| Mask3D (MAE)Backbone=ViT, Pre-training Data=ImageNet+ScanNet2023.02 | 66.7 | |
| Supervised Pre-trainingBackbone=ViT, Pre-training Data=ImageNet+ScanNet2023.02 | 65.9 | |
| MAE-unsupIN→SNBackbone=ViT, Pre-training Data=ImageNet+ScanNet2023.02 | 63.3 | |
| Mask3D (DINO)Backbone=ViT, Pre-training Data=ImageNet+ScanNet2023.02 | 60.5 | |
| Pri3DBackbone=ResNet-50, Pre-training Data=ImageNet+ScanNet2023.02 | 60.2 | |
| Pri3DBackbone=ViT, Pre-training Data=ImageNet+ScanNet2023.02 | 59.3 | |
| DINOBackbone=ViT, Pre-training Data=ImageNet+ScanNet2023.02 | 58.1 | |
| MoCoV2-supIN→SNBackbone=ResNet-50, Pre-training Data=ImageNet+ScanNet2023.02 | 56.6 | |
| ImageNet Pre-training (supIN)Backbone=ResNet-50, Pre-training Data=ImageNet2023.02 | 55.7 | |
| ScratchBackbone=ResNet-50, Pre-training Data=None2023.02 | 39.1 |