Instance Segmentation on COCO 2017 (val)
50.3APmSwin-Small
Evaluation results
| Method | Links | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Swin-SmallBackbone=Swin-Small, Segmentation Head=Mask2Former, Search Space=100 queries, Epochs=50, # Params=69M, FLOPS=313G2023.04 | 50.3 | — | — | — | — | 46.3 | 25.3 | — | 68.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AFF-SmallBackbone=AFF-Small, Segmentation Head=Mask2Former*, Search Space=100 queries, Epochs=50, # Params=61.4M, FLOPS=281G2023.04 | 49.8 | — | — | — | — | 46.4 | 27 | — | 67.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AFF-TinyBackbone=AFF-Tiny, Segmentation Head=Mask2Former*, Search Space=100 queries, Epochs=50, # Params=46M, FLOPS=204G2023.04 | 49.2 | — | — | — | — | 45.3 | 24.8 | — | 66.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AFF-Small-1/5Backbone=AFF-Small-1/5, Segmentation Head=Mask2Former*, Search Space=100 queries, Epochs=50, # Params=61.4M, FLOPS=206G2023.04 | 49.2 | — | — | — | — | 45.7 | 26.1 | — | 67.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-TinyBackbone=Swin-Tiny, Segmentation Head=Mask2Former, Search Space=100 queries, Epochs=50, # Params=47M, FLOPS=232G2023.04 | 48.3 | — | — | — | — | 45 | 24.5 | — | 67.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AFF-Tiny-1/5Backbone=AFF-Tiny-1/5, Segmentation Head=Mask2Former*, Search Space=100 queries, Epochs=50, # Params=46M, FLOPS=152G2023.04 | 47.8 | — | — | — | — | 44.5 | 24.5 | — | 66.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AFF-Mini-1/5Backbone=AFF-Mini-1/5, Segmentation Head=Mask2Former*, Search Space=100 queries, Epochs=50, # Params=25.8M, FLOPS=120G2023.04 | 45.7 | — | — | — | — | 42.3 | 21.8 | — | 64 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| AFF-MiniBackbone=AFF-Mini, Segmentation Head=Mask2Former*, Search Space=100 queries, Epochs=50, # Params=25.8M, FLOPS=148G2023.04 | 45.6 | — | — | — | — | 42.3 | 21.2 | — | 63.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| UniFormer-B#Params (M)=69, FLOPs (G)=399, Schedule=3× + MS, Pre-training=ImageNet-1K, Resolution=800×1280, Detection Framework=Mask R-CNN, Window Size=142022.01 | 44.8 | — | — | 69 | 48.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| UniFormer-S#Params (M)=41, FLOPs (G)=269, Schedule=3× + MS, Pre-training=ImageNet-1K, Resolution=800×1280, Detection Framework=Mask R-CNN, Window Size=142022.01 | 43.4 | — | — | 67.1 | 47 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-SModel Scale=Small, Params (M)=69, Schedule=3x2022.11 | 43.3 | — | — | 67.3 | 46.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| UniFormer-B#Params (M)=69, FLOPs (G)=399, Schedule=1×, Pre-training=ImageNet-1K, Resolution=800×1280, Detection Framework=Mask R-CNN, Window Size=142022.01 | 43.1 | — | — | 66 | 46.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SALG-TModel Scale=Tiny, Params (M)=52, Schedule=3x2022.11 | 42.2 | — | — | 65.9 | 45.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SALG-SModel Scale=Small, Params (M)=73, Schedule=3x2022.11 | 41.7 | — | — | 65.5 | 44.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-TModel Scale=Tiny, Params (M)=48, Schedule=3x2022.11 | 41.6 | — | — | 65.1 | 44.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| UniFormer-S#Params (M)=41, FLOPs (G)=269, Schedule=1×, Pre-training=ImageNet-1K, Resolution=800×1280, Detection Framework=Mask R-CNN, Window Size=142022.01 | 41.6 | — | — | 64.8 | 45 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CONTAINER-LIGHTFramework=Mask R-CNN, #P=39.6, FLOPs=237G2021.06 | 41.3 | — | — | 64.2 | 44.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ViL-BFramework=Mask R-CNN, #P=76.1, FLOPs=365G2021.06 | 41 | — | — | 64.3 | 44.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-SModel Scale=Small, Params (M)=69, Schedule=1x2022.11 | 40.9 | — | — | 63.6 | 44.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PVT-MModel Scale=Small, Params (M)=64, Schedule=3x2022.11 | 40.5 | — | — | 63.1 | 43.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SALG-SModel Scale=Small, Params (M)=73, Schedule=1x2022.11 | 40.4 | — | — | 64.2 | 43.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SALG-TModel Scale=Tiny, Params (M)=52, Schedule=1x2022.11 | 40 | — | — | 63.8 | 42.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PVT-SModel Scale=Tiny, Params (M)=44, Schedule=3x2022.11 | 39.9 | — | — | 62.5 | 42.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SWIN-TFramework=Mask R-CNN, #P=48.0, FLOPs=267G2021.06 | 39.8 | — | — | 63.6 | 42.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ViL-MFramework=Mask R-CNN, #P=60.1, FLOPs=261G2021.06 | 39.7 | — | — | 62.8 | 42.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-TModel Scale=Tiny, Params (M)=48, Schedule=1x2022.11 | 39.1 | — | — | 61.6 | 42 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PVT-MModel Scale=Small, Params (M)=64, Schedule=1x2022.11 | 39 | — | — | 61.6 | 42.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BoT50-(6x)Framework=Mask R-CNN, #P=39.52021.06 | 38.7 | — | — | 61.8 | 41.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ViL-SFramework=Mask R-CNN, #P=45.0, FLOPs=174G2021.06 | 38.5 | — | — | 61.1 | 41.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Res101Model Scale=Small, Params (M)=63, Schedule=3x2022.11 | 38.5 | — | — | 60.1 | 41.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| X-101-64Framework=Mask R-CNN, #P=101.9, FLOPs=424G2021.06 | 38.4 | — | — | 60.6 | 41.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SALG-STModel Scale=Super-tiny, Params (M)=26, Schedule=3x2022.11 | 38.2 | — | — | 60.7 | 40.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PVT-SFramework=Mask R-CNN, #P=44.1, FLOPs=245G2021.06 | 37.8 | — | — | 60.1 | 40.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PVT-SModel Scale=Tiny, Params (M)=44, Schedule=1x2022.11 | 37.8 | — | — | 60.1 | 40.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| X-101-32Framework=Mask R-CNN, #P=62.8, FLOPs=259G2021.06 | 37.5 | — | — | 59.4 | 40.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Res50Model Scale=Tiny, Params (M)=44, Schedule=3x2022.11 | 37.1 | — | — | 58.4 | 40.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Res50#Params (M)=44, FLOPs (G)=260, Schedule=3× + MS, Pre-training=ImageNet-1K, Resolution=800×1280, Detection Framework=Mask R-CNN2022.01 | 37.1 | — | — | 58.4 | 40.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Res101Model Scale=Small, Params (M)=63, Schedule=1x2022.11 | 36.4 | — | — | 57.7 | 38.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-STModel Scale=Super-tiny, Params (M)=25, Schedule=3x2022.11 | 36.3 | — | — | 57.6 | 39 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ResNet101Framework=Mask R-CNN, #P=63.2, FLOPs=259G2021.06 | 36.1 | — | — | 57.5 | 38.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BoT50Framework=Mask R-CNN, #P=39.52021.06 | 35.3 | — | — | 57 | 37.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SALG-STModel Scale=Super-tiny, Params (M)=26, Schedule=1x2022.11 | 35.3 | — | — | 57.6 | 37.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-Mini+Backbone=Swin-Mini+, Segmentation Head=Mask2Former*, Search Space=100 queries, Epochs=50, # Params=25.8M, FLOPS=149G2023.04 | 35.2 | — | — | — | — | 33.1 | 13.8 | — | 53.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ResNet50Framework=Mask R-CNN, #P=44.2, FLOPs=180G2021.06 | 34.7 | — | — | 55.7 | 37.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Res50Model Scale=Tiny, Params (M)=44, Schedule=1x2022.11 | 34.4 | — | — | 55.1 | 36.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Res50#Params (M)=44, FLOPs (G)=260, Schedule=1×, Pre-training=ImageNet-1K, Resolution=800×1280, Detection Framework=Mask R-CNN2022.01 | 34.4 | — | — | 55.1 | 36.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-STModel Scale=Super-tiny, Params (M)=25, Schedule=1x2022.11 | 32.4 | — | — | 53.4 | 33.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Cascade MRCNN-RSBackbone=ResNet152-FPN, Resolution=1280x1280, Latency (ms) FP16=119, Latency (ms) FP32=1912021.06 | 0.567 | — | — | — | — | 52.9 | 33.5 | — | 70.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Cascade MRCNN-RSBackbone=SpineNet143L, Resolution=1280x1280, Latency (ms) FP16=144, Latency (ms) FP32=2342021.06 | 0.567 | — | — | — | — | 53.6 | 34.5 | — | 70.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Cascade MRCNN-RSBackbone=ResNet200-FPN, Resolution=1280x1280, Latency (ms) FP16=149, Latency (ms) FP32=2322021.06 | 0.562 | — | — | — | — | 53.1 | 33.9 | — | 70.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Cascade MRCNN-RSBackbone=EfficientNetB7-FPN, Resolution=1536x1536, Latency (ms) FP16=149, Latency (ms) FP32=2632021.06 | 0.558 | — | — | — | — | 52.6 | 32.2 | — | 70.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Cascade MRCNN-RSBackbone=SpineNet143, Resolution=1280x1280, Latency (ms) FP16=109, Latency (ms) FP32=1752021.06 | 0.558 | — | — | — | — | 53 | 33.8 | — | 70.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Cascade MRCNN-RSBackbone=EfficientNetB6-FPN, Resolution=1280x1280, Latency (ms) FP16=101, Latency (ms) FP32=1792021.06 | 0.554 | — | — | — | — | 52.5 | 33 | — | 69.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Cascade MRCNN-RSBackbone=EfficientNetB5-FPN, Resolution=1280x1280, Latency (ms) FP16=90, Latency (ms) FP32=1582021.06 | 0.553 | — | — | — | — | 52 | 32.3 | — | 69.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Cascade MRCNN-RSBackbone=ResNet152-FPN, Resolution=1024x1024, Latency (ms) FP16=90, Latency (ms) FP32=1482021.06 | 0.553 | — | — | — | — | 52.4 | 32.9 | — | 70 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Cascade MRCNN-RSBackbone=ResNet152-FPN, Resolution=896x896, Latency (ms) FP16=79, Latency (ms) FP32=1252021.06 | 0.551 | — | — | — | — | 51.8 | 32 | — | 70 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Cascade MRCNN-RSBackbone=ResNet101-FPN, Resolution=896x896, Latency (ms) FP16=68, Latency (ms) FP32=1092021.06 | 0.543 | — | — | — | — | 51.1 | 31.4 | — | 69.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Cascade MRCNN-RSBackbone=ResNet101-FPN, Resolution=768x768, Latency (ms) FP16=60, Latency (ms) FP32=972021.06 | 0.539 | — | — | — | — | 50.3 | 29.7 | — | 69.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RefineMaskBackbone=X101-FPN, Schedule=3x, Multi-scale training=true2021.04 | 0.531 | — | — | — | — | 41.5 | 28.6 | — | 62.8 | — | — | — | — | — | — | — | — | — | — | — | — | 45.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Cascade MRCNN-RSBackbone=EfficientNetB4-FPN, Resolution=1024x1024, Latency (ms) FP16=67, Latency (ms) FP32=1092021.06 | 0.531 | — | — | — | — | 50.5 | 30.5 | — | 69.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Cascade MRCNN-RSBackbone=ResNet50-FPN, Resolution=768x768, Latency (ms) FP16=50, Latency (ms) FP32=852021.06 | 0.526 | — | — | — | — | 49.4 | 29.4 | — | 68.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Cascade MRCNN-RSBackbone=EfficientNetB3-FPN, Resolution=896x896, Latency (ms) FP16=54, Latency (ms) FP32=902021.06 | 0.52 | — | — | — | — | 48.9 | 29.2 | — | 67.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Cascade MRCNN-RSBackbone=ResNet50-FPN, Resolution=640x640, Latency (ms) FP16=45, Latency (ms) FP32=772021.06 | 0.515 | — | — | — | — | 48.4 | 27.4 | — | 67.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ViTDet-LFramework=Cascade Mask R-CNN, Epoch=100, Pre-train=IN-1K, MAE2022.05 | 0.511 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 52.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ViT-Adapter-LFramework=HTC++, Epoch=36, Pre-train=MM, BEiTv22022.05 | 0.511 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 52.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CBV2-Swin-LFramework=HTC, Epoch=36, Pre-train=IN-22K, supervised2022.05 | 0.51 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 51.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Mask2former + DFPQBackbone=Swin-B, Epochs=36, Parameters (M)=115, FLOPs (G)=4662022.04 | 0.51 | — | — | — | — | 47.4 | 26.8 | — | 70.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ViT-Adapter-LFramework=HTC++, Epoch=36, Pre-train=IN-22K, BEIT2022.05 | 0.508 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 52.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Mask2formerBackbone=Swin-B, Epochs=36, Parameters (M)=115, FLOPs (G)=4662022.04 | 0.508 | — | — | — | — | 47.2 | 27.1 | — | 69.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RefineMaskBackbone=R101-FPN, Schedule=2x2021.04 | 0.502 | — | — | — | — | 38.8 | 24.7 | — | 61.7 | — | — | — | — | — | — | — | — | — | — | — | — | 42.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Cascade MRCNN-RSBackbone=ResNet50-FPN, Resolution=512x512, Latency (ms) FP16=41, Latency (ms) FP32=692021.06 | 0.501 | — | — | — | — | 46.1 | 23.7 | — | 67.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Focal-LFramework=HTC++, Epoch=36, Pre-train=IN-22K, supervised2022.05 | 0.499 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 50.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Cascade MRCNN-RSBackbone=EfficientNetB2-FPN, Resolution=768x768, Latency (ms) FP16=48, Latency (ms) FP32=782021.06 | 0.498 | — | — | — | — | 46.9 | 26.4 | — | 66.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-LFramework=HTC++, Epoch=36, Pre-train=IN-1K, UM-MAE2022.05 | 0.498 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 50.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Mask2FormerAnn.=M, Backbone=Swin-S, sched.=50e, Arch.=Mask2Former2023.12 | 0.497 | 69.4 | 52.8 | — | — | 46.1 | 25.4 | — | 68.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RefineMaskBackbone=R101-FPN, Schedule=1x2021.04 | 0.495 | — | — | — | — | 38.6 | 24.9 | — | 59.9 | — | — | — | — | — | — | — | — | — | — | — | — | 41.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Swin-LFramework=HTC++, Epoch=72, Pre-train=IN-22K, supervised2022.05 | 0.495 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 50.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Mask R-CNNBackbone=X101-FPN, Schedule=3x2021.04 | 0.49 | — | — | — | — | 39.4 | 27.2 | — | 57.7 | — | — | — | — | — | — | — | — | — | — | — | — | 41.8 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ViT-Adapter-LFramework=HTC++, Epoch=36, Pre-train=IN-22K, supervised2022.05 | 0.49 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 49.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ConvNeXt V2-HMethod=FCMAE, FLOPS=2525G2023.01 | 0.489 | — | — | 72.8 | 53.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MogaNet-XL Cascade Mask R-CNNBackbone=MogaNet-XL, Framework=Cascade Mask R-CNN, Schedule=multi-scale 3x, Pre-training=IN-21K, Resolution=800x1280, Parameters=238M, FLOPs=1355G2022.11 | 0.488 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MogaNet-XLType=Convolutional, #P (M)=238, FLOPs (G)=1355, Pre-training=ImageNet-21K, Framework=Cascade Mask R-CNN, Training Schedule=3x, Multi-scaling=true, Resolution=800x12802022.11 | 0.488 | 72.6 | 53.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RefineMaskBackbone=R50-FPN, Schedule=1x2021.04 | 0.488 | — | — | — | — | 37.3 | 24.1 | — | 58 | — | — | — | — | — | — | — | — | — | — | — | — | 40.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Cascade MRCNN-RSBackbone=EfficientNetB1-FPN, Resolution=640x640, Latency (ms) FP16=42, Latency (ms) FP32=712021.06 | 0.488 | — | — | — | — | 45.5 | 24.3 | — | 65 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MViTv2-HFramework=Cascade Mask R-CNN, Epoch=50, Pre-train=IN-22K, supervised2022.05 | 0.488 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 50.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| InternImage-XL+detector=Cascade Mask R-CNN, schedule=3x, MS=true, parameters=387M, FLOPs=1782G, Input Resolution=1280x8002022.11 | 0.488 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 72.5 | 53.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RefineMaskBackbone=R50-FPN, Schedule=2x2021.04 | 0.487 | — | — | — | — | 37.9 | 24.5 | — | 59.8 | — | — | — | — | — | — | — | — | — | — | — | — | 41.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| HorNet-L Cascade Mask R-CNNBackbone=HorNet-L, Framework=Cascade Mask R-CNN, Schedule=multi-scale 3x, Pre-training=IN-21K, Resolution=800x1280, Parameters=259M, FLOPs=1399G2022.11 | 0.486 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| HorNet-LType=Convolutional, #P (M)=259, FLOPs (G)=1399, Pre-training=ImageNet-21K, Framework=Cascade Mask R-CNN, Training Schedule=3x, Multi-scaling=true, Resolution=800x12802022.11 | 0.486 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MViTv2-LFramework=Cascade Mask R-CNN, Epoch=50, Pre-train=IN-22K, supervised2022.05 | 0.486 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 50.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| HorNet-L+detector=Cascade Mask R-CNN, schedule=3x, MS=true, parameters=259M, FLOPs=1358G, Input Resolution=1280x8002022.11 | 0.486 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| InternImage-L+detector=Cascade Mask R-CNN, schedule=3x, MS=true, parameters=277M, FLOPs=1399G, Input Resolution=1280x8002022.11 | 0.485 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 72.4 | 53 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| InternImage-XL+detector=Cascade Mask R-CNN, schedule=1x, MS=false, parameters=387M, FLOPs=1782G, Input Resolution=1280x8002022.11 | 0.481 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 71.9 | 52.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RAVLT-LType=Linear, Params (M)=152, FLOPs (G)=979, Detector=Cascade Mask R-CNN, Training Schedule=3x + MS2024.11 | 0.48 | — | — | 71.8 | 52.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DINAT-LBackbone=DINAT-L, Detection Framework=Cascade Mask R-CNN, Training Schedule=3x, Number of Parameters=258 M, FLOPs=1276 G, Pre-training (ImageNet-22K)=true2022.09 | 0.478 | — | — | 71.8 | 52 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 14 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DINAT-L+Framework=Cascade Mask R-CNN, Schedule=3x, Params=258 M, FLOPs=1276 G, Throughput=14.0 FPS, Pre-trained=ImageNet-22K2022.09 | 0.478 | — | — | 71.8 | 52 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ConvNeXt-XLType=Convolutional, #P (M)=407, FLOPs (G)=1898, Pre-training=ImageNet-21K, Framework=Cascade Mask R-CNN, Training Schedule=3x, Multi-scaling=true, Resolution=800x12802022.11 | 0.477 | 71.6 | 52.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PIIP-SBLFramework=Cascade R-CNN, Training Schedule=3x, Multi-scale training=true, Pre-training=ImageNet-21K2025.01 | 0.477 | — | — | 71.6 | 52.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Mask2Formersupervision=mask-supervised, backbone=ResNet-1012022.12 | 0.477 | — | — | — | — | 44.2 | 23.8 | — | 66.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ConvNeXt V2-LMethod=FCMAE, FLOPS=875G2023.01 | 0.477 | — | — | 71.4 | 52.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — |