Instance Segmentation on MS COCO 2014 (val)
54.14AP^mDense
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| DenseModel=ViT-H, PR (%)=0, Box prompt generator=FocalNet-DINO, Evaluation Protocol=Zero-shot segmentation2026.02 | 54.14 | 75.2 | 53.23 | 49.03 | 37.13 | 61.78 | 55.09 | |
| DenseModel=ViT-L, PR (%)=0, Box prompt generator=FocalNet-DINO, Evaluation Protocol=Zero-shot segmentation2026.02 | 53.47 | 74.97 | 52.94 | 48.72 | 36.85 | 61.9 | 54.81 | |
| RandomModel=ViT-H, PR (%)=20, Box prompt generator=FocalNet-DINO, Evaluation Protocol=Zero-shot segmentation2026.02 | 52.82 | 73.98 | 51.68 | 47.7 | 35.52 | 61.25 | 53.82 | |
| POPModel=ViT-H, PR (%)=20, Box prompt generator=FocalNet-DINO, Evaluation Protocol=Zero-shot segmentation2026.02 | 52.08 | 74.23 | 50.99 | 47.56 | 36.09 | 60.32 | 53.54 | |
| RandomModel=ViT-L, PR (%)=20, Box prompt generator=FocalNet-DINO, Evaluation Protocol=Zero-shot segmentation2026.02 | 51.96 | 73.66 | 50.7 | 47.18 | 35.01 | 60.55 | 53.18 | |
| POPModel=ViT-L, PR (%)=20, Box prompt generator=FocalNet-DINO, Evaluation Protocol=Zero-shot segmentation2026.02 | 51.71 | 74.14 | 50.68 | 47.23 | 35.88 | 59.93 | 53.26 | |
| DenseModel=ViT-B, PR (%)=0, Box prompt generator=FocalNet-DINO, Evaluation Protocol=Zero-shot segmentation2026.02 | 50.52 | 73.17 | 49.21 | 46.16 | 34.61 | 59.32 | 52.16 | |
| POPModel=ViT-B, PR (%)=20, Box prompt generator=FocalNet-DINO, Evaluation Protocol=Zero-shot segmentation2026.02 | 49.57 | 72.79 | 47.94 | 45.38 | 34.06 | 58.32 | 51.34 | |
| RandomModel=ViT-B, PR (%)=20, Box prompt generator=FocalNet-DINO, Evaluation Protocol=Zero-shot segmentation2026.02 | 48.66 | 71.76 | 46.36 | 44.31 | 32.02 | 57.86 | 50.16 | |
| DAT-TDecoder=Cascade Mask RCNN, Schedule=3x, Params (M)=86, Venue=CVPR'222022.02 | 44.5 | 67.5 | 48.1 | — | — | — | — | |
| ViTAEv2-SDecoder=Cascade Mask RCNN, Schedule=3x, Params (M)=752022.02 | 44.5 | 67.8 | 48.2 | — | — | — | — | |
| Swin-TDecoder=Cascade Mask RCNN, Schedule=3x, Params (M)=86, Venue=ICCV'212022.02 | 43.7 | 66.6 | 47.3 | — | — | — | — | |
| ViTAEv2-SDecoder=Cascade Mask RCNN, Schedule=1x, Params (M)=752022.02 | 43.6 | 66.9 | 47.2 | — | — | — | — | |
| ViTAEv2-SDecoder=Mask RCNN, Schedule=3x, Params (M)=372022.02 | 42.6 | 66.6 | 45.8 | — | — | — | — | |
| DAT-TDecoder=Cascade Mask RCNN, Schedule=1x, Params (M)=86, Venue=CVPR'222022.02 | 42.5 | 65.4 | 45.8 | — | — | — | — | |
| DAT-TDecoder=Mask RCNN, Schedule=3x, Params (M)=48, Venue=CVPR'222022.02 | 42.4 | 66.1 | 45.5 | — | — | — | — | |
| MViT-TDecoder=Mask RCNN, Schedule=3x, Params (M)=46, Venue=ICCV'212022.02 | 42.1 | 66 | 45.4 | — | — | — | — | |
| ViTAEv2-SDecoder=Mask RCNN, Schedule=1x, Params (M)=372022.02 | 41.8 | 65.6 | 44.9 | — | — | — | — | |
| Swin-TDecoder=Cascade Mask RCNN, Schedule=1x, Params (M)=86, Venue=ICCV'212022.02 | 41.7 | 64.4 | 45 | — | — | — | — | |
| Swin-TDecoder=Mask RCNN, Schedule=3x, Params (M)=48, Venue=ICCV'212022.02 | 41.6 | 65.1 | 44.8 | — | — | — | — | |
| CrossFormer-SDecoder=Mask RCNN, Schedule=1x, Params (M)=50, Venue=ICLR'212022.02 | 41.4 | 64.8 | 44.6 | — | — | — | — | |
| PVTv2-B2Decoder=Mask RCNN, Schedule=1x, Params (M)=45, Venue=CVMJ'222022.02 | 41.2 | 64.2 | 44.4 | — | — | — | — | |
| Focal-TDecoder=Mask RCNN, Schedule=1x, Params (M)=49, Venue=NeurIPS'212022.02 | 41 | 64.7 | 44.2 | — | — | — | — | |
| RegionViT-BDecoder=Mask RCNN, Schedule=1x, Params (M)=92, Venue=ICLR'212022.02 | 40.8 | 63.4 | 44 | — | — | — | — | |
| DAT-TDecoder=Mask RCNN, Schedule=1x, Params (M)=48, Venue=CVPR'222022.02 | 40.4 | 64.2 | 43.1 | — | — | — | — | |
| ResNet-50Decoder=Cascade Mask RCNN, Schedule=3x, Params (M)=82, Venue=CVPR'162022.02 | 40.1 | 61.7 | 43.4 | — | — | — | — | |
| PVT-SDecoder=Mask RCNN, Schedule=3x, Params (M)=44, Venue=ICCV'212022.02 | 39.9 | 62.5 | 42.8 | — | — | — | — | |
| Swin-TDecoder=Mask RCNN, Schedule=1x, Params (M)=47, Venue=ICCV'212022.02 | 39.8 | 63.3 | 42.7 | — | — | — | — | |
| Conformer-S/32Decoder=Mask RCNN, Schedule=1x, Params (M)=58, Venue=ICCV'212022.02 | 39.7 | — | — | — | — | — | — | |
| PVT-SDecoder=Mask RCNN, Schedule=1x, Params (M)=44, Venue=ICCV'212022.02 | 37.8 | 60.1 | 40.3 | — | — | — | — | |
| ResNet-50Decoder=Mask RCNN, Schedule=3x, Params (M)=44, Venue=CVPR'162022.02 | 37.1 | 58.4 | 40.1 | — | — | — | — | |
| ResNet-50Decoder=Cascade Mask RCNN, Schedule=1x, Params (M)=82, Venue=CVPR'162022.02 | 35.9 | 56.6 | 38.4 | — | — | — | — | |
| ResNet-50Decoder=Mask RCNN, Schedule=1x, Params (M)=44, Venue=CVPR'162022.02 | 34.4 | 55.1 | 36.7 | — | — | — | — |