Semantic Segmentation on ADE20K 150 categories (test)
79.36Pixel AccuracyMaskFormer
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| MaskFormerbackbone=Swin-L, Pre-trained=ImageNet-22K [29], Training set=ADE20K train + val, Multi-scale inference=true2021.07 | 79.36 | 49.67 | 64.51 | |
| MaskFormerbackbone=Swin-L, Pre-trained=ImageNet-22K, Training set=ADE20K train + val, Multi-scale inference=true2021.07 | 79.36 | 49.67 | 64.51 | |
| Swin-UperNetbackbone=Swin-L2021.07 | 78.42 | 47.07 | 62.75 | |
| Swin-UperNetbackbone=Swin-L2021.07 | 78.42 | 47.07 | 62.75 | |
| SETRbackbone=ViT-L2021.07 | 78.35 | 45.03 | 61.69 | |
| SETRbackbone=ViT-L2021.07 | 78.35 | 45.03 | 61.69 |