Instance Segmentation on PASCAL VOC 2012 (val)
77.6mAP @0.5APro
Evaluation Results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| AProFramework=Mask2Former, Backbone=Swin-L, #Epoch=502023.10 | 77.6 | — | — | — | — | — | — | — | 53.1 | 49.6 | — | |
| AProFramework=Mask2Former, Backbone=ResNet-101, #Epoch=502023.10 | 72 | — | — | — | — | — | — | — | 45.7 | 43.6 | — | |
| AProFramework=Mask2Former, Backbone=ResNet-50, #Epoch=502023.10 | 70.6 | — | — | — | — | — | — | — | 44.5 | 42.3 | — | |
| Mask R-CNNSupervision=Full supervision: Instance masks2021.03 | 69.1 | 49.9 | — | — | — | — | 65.8 | 77.3 | 41.9 | — | — | |
| MRCNNSupervision=segmentation label, Extra data / Information=MS-COCO [31]2019.04 | 69 | — | — | — | — | — | — | — | — | — | — | |
| Mask R-CNNSupervision=Fully supervised, Backbone=ResNet-1012020.07 | 67.9 | 52.5 | — | — | — | — | — | 76.7 | 44.9 | — | — | |
| Mask R-CNNSupervision source=Full Mask (F)2021.09 | 67.9 | 52.5 | — | — | — | — | — | 76.7 | 44.9 | — | — | |
| AProFramework=SOLOv2, Backbone=ResNet-101, #Epoch=362023.10 | 67.9 | — | — | — | — | — | — | — | 42.6 | 40.5 | — | |
| Mask R-CNNBackbone=ResNet-101, Supervision=Instance-level (M), Refinement=false2024.02 | 67.9 | 52.5 | — | — | — | — | — | 76.7 | 44.9 | — | — | |
| BoxTeacherBackbone=ResNet-1012022.10 | 67.8 | 48 | — | — | — | — | — | 78.4 | 41.3 | 40.3 | — | |
| SIMbackbone=ResNet-1012023.03 | 67.1 | — | — | — | — | — | — | — | 38.3 | 38.6 | — | |
| BoxTeacherBackbone=ResNet-502022.10 | 66.4 | 46.1 | — | — | — | — | — | 77.6 | 38.7 | 38.6 | — | |
| BoxLevelSetbackbone=ResNet-1012023.03 | 66.3 | — | — | — | — | — | — | — | 38.7 | 38.3 | — | |
| BoxLevelSetBackbone=ResNet-1012022.10 | 66.3 | 46.4 | — | — | — | — | — | 77.9 | 38.7 | 38.3 | — | |
| SIMbackbone=ResNet-502023.03 | 65.5 | — | — | — | — | — | — | — | 35.6 | 36.7 | — | |
| AProFramework=SOLOv2, Backbone=ResNet-50, #Epoch=362023.10 | 65.4 | — | — | — | — | — | — | — | 39.8 | 38.4 | — | |
| AProFramework=SOLOv2, Backbone=ResNet-50, #Epoch=122023.10 | 65.1 | — | — | — | — | — | — | — | 37 | 37.1 | — | |
| AProFramework=Mask2Former, Backbone=ResNet-50, #Epoch=122023.10 | 65.1 | — | — | — | — | — | — | — | 37 | 37 | — | |
| TreeEnergy LossFramework=Mask2Former, Backbone=ResNet-50, #Epoch=122023.10 | 65 | — | — | — | — | — | — | — | 34.3 | 36 | — | |
| CRF LossFramework=SOLOv2, Backbone=ResNet-50, #Epoch=122023.10 | 64.7 | — | — | — | — | — | — | — | 34.9 | 35 | — | |
| TreeEnergy LossFramework=SOLOv2, Backbone=ResNet-50, #Epoch=122023.10 | 64.4 | — | — | — | — | — | — | — | 34.7 | 35 | — | |
| Pairwise LossFramework=SOLOv2, Backbone=ResNet-50, #Epoch=122023.10 | 64.3 | — | — | — | — | — | — | — | 35.1 | 35.7 | — | |
| CRF LossFramework=Mask2Former, Backbone=ResNet-50, #Epoch=122023.10 | 64.3 | — | — | — | — | — | — | — | 35.2 | 35.7 | — | |
| BoxLevelSetbackbone=ResNet-502023.03 | 64.2 | — | — | — | — | — | — | — | 35.9 | 36.3 | — | |
| BoxLevelSetBackbone=ResNet-502022.10 | 64.2 | 43.9 | — | — | — | — | — | 76.3 | 35.9 | 36.3 | — | |
| BoxLevelsetBackbone=ResNet-50, #Epoch=362023.10 | 64.2 | — | — | — | — | — | — | — | 35.9 | 36.3 | — | |
| CRF LossFramework=SOLOv2, Backbone=ResNet-50, #Epoch=362023.10 | 64 | — | — | — | — | — | — | — | 35.7 | 35.9 | — | |
| BBAMSupervision=Weak supervision: Bounding boxes2021.03 | 63.7 | 39.5 | — | — | — | — | 63 | 76.8 | 31.8 | — | — | |
| BBAMbackbone=ResNet-1012023.03 | 63.7 | — | — | — | — | — | — | — | 31.8 | — | — | |
| BBAMBackbone=ResNet-1012022.10 | 63.7 | 39.5 | — | — | — | — | — | 76.8 | 31.8 | — | — | |
| DISCOBOXBackbone=ResNet-50-DCN, Arch=YOLACT++2021.05 | 63.6 | 41.6 | — | — | — | — | — | 75.2 | 34.1 | — | — | |
| MNCbackbone=VGG-16, time/img (s)=0.362015.12 | 63.5 | 41.5 | — | — | — | — | — | — | — | — | — | |
| MNC2016.05 | 63.5 | 41.5 | — | — | — | — | — | — | — | — | — | |
| MNC2016.11 | 63.5 | 41.5 | — | — | — | — | — | — | — | — | — | |
| MNCSupervision=Full supervision: Instance masks2021.03 | 63.5 | 41.5 | — | — | — | — | — | — | — | — | — | |
| TreeEnergy LossFramework=SOLOv2, Backbone=ResNet-50, #Epoch=362023.10 | 63.5 | — | — | — | — | — | — | — | 36.1 | 36.1 | — | |
| Pairwise LossFramework=SOLOv2, Backbone=ResNet-50, #Epoch=362023.10 | 63.4 | — | — | — | — | — | — | — | 38.1 | 36.5 | — | |
| Pairwise LossFramework=Mask2Former, Backbone=ResNet-50, #Epoch=122023.10 | 62.9 | — | — | — | — | — | — | — | 33.9 | 35.2 | — | |
| DISCOBOXBackbone=ResNet-101, Arch=SOLOv22021.05 | 62.2 | 45.5 | — | — | — | — | — | 72.8 | 37.5 | — | — | |
| DiscoBoxbackbone=ResNet-1012023.03 | 62.2 | — | — | — | — | — | — | — | 37.5 | — | — | |
| DiscoBoxBackbone=ResNet-1012022.10 | 62.2 | 45.5 | — | — | — | — | — | 72.8 | 37.5 | — | — | |
| MPA2016.11 | 61.8 | — | — | — | — | — | — | — | — | — | — | |
| FCRNLayer depth=101, Bootstrapping=true, Loss function=weighted loss, Additional training data=COCO2016.05 | 61.5 | 46.6 | 56.4 | — | — | — | — | — | — | — | — | |
| BoxInstbackbone=ResNet-1012020.12 | 61.4 | — | — | — | — | — | — | — | 37 | 36.5 | — | |
| BoxInstBackbone=ResNet-101, Arch=CondInst2021.05 | 61.4 | — | — | — | — | — | — | — | 37 | — | — | |
| BoxInstbackbone=ResNet-1012023.03 | 61.4 | — | — | — | — | — | — | — | 37 | 36.5 | — | |
| BoxInstBackbone=ResNet-1012022.10 | 61.4 | — | — | — | — | — | — | — | 37 | 36.4 | — | |
| FCRNLayer depth=101, Bootstrapping=true, Loss function=weighted loss2016.05 | 60.9 | 44.6 | 55.5 | — | — | — | — | — | — | — | — | |
| CFMbackbone=VGG-16, proposal=MCG, time/img (s)=322015.12 | 60.7 | 39.6 | — | — | — | — | — | — | — | — | — | |
| CFMBackbone=VGG, Proposals=MCG2014.12 | 60.7 | — | — | — | — | — | — | — | — | — | — | |
| CFM2016.11 | 60.7 | 39.6 | — | — | — | — | — | — | — | — | — | |
| Hypercolumnbackbone=VGG-16, proposal=MCG, time/img (s)=>802015.12 | 60 | 40.4 | — | — | — | — | — | — | — | — | — | |
| Hypercolumn2016.05 | 60 | 40.4 | — | — | — | — | — | — | — | — | — | |
| Hypercolumn2016.11 | 60 | 40.4 | — | — | — | — | — | — | — | — | — | |
| FCRNLayer depth=50, Bootstrapping=true, Loss function=weighted loss2016.05 | 59.8 | 43.5 | 54.7 | — | — | — | — | — | — | — | — | |
| DISCOBOXBackbone=ResNet-50, Arch=SOLOv22021.05 | 59.8 | 41.7 | — | — | — | — | — | 71.4 | 35.5 | — | — | |
| DiscoBoxbackbone=ResNet-502023.03 | 59.8 | — | — | — | — | — | — | — | 35.5 | — | — | |
| DiscoBoxBackbone=ResNet-502022.10 | 59.8 | 41.7 | — | — | — | — | — | 71.4 | 35.5 | — | — | |
| DiscoBoxBackbone=ResNet-50, #Epoch=362023.10 | 59.8 | — | — | — | — | — | — | — | 35.5 | — | — | |
| BBTP w/ CRFbackbone=ResNet-1012020.12 | 59.1 | — | — | — | — | — | — | — | 21.9 | 27.5 | — | |
| BoxInstbackbone=ResNet-502020.12 | 59.1 | — | — | — | — | — | — | — | 34.2 | 34.3 | — | |
| BoxInstBackbone=ResNet-50, Arch=CondInst2021.05 | 59.1 | — | — | — | — | — | — | — | 34.2 | — | — | |
| BBTP w/ CRFbackbone=ResNet-1012023.03 | 59.1 | — | — | — | — | — | — | — | 21.9 | 27.5 | — | |
| BoxInstbackbone=ResNet-502023.03 | 59.1 | — | — | — | — | — | — | — | 34.2 | 34.3 | — | |
| BoxInstBackbone=ResNet-502022.10 | 59.1 | — | — | — | — | — | — | — | 34.2 | 34.3 | — | |
| BBTPSupervision=Bounding box, Backbone=ResNet-1012020.07 | 58.9 | 30.4 | — | — | — | — | — | 75 | 21.6 | — | — | |
| Hsu et al.Supervision=Weak supervision: Bounding boxes2021.03 | 58.9 | 30.4 | — | — | — | — | — | 75 | 21.6 | — | — | |
| BBTPBackbone=ResNet-101, Arch=Mask R-CNN2021.05 | 58.9 | 30.4 | — | — | — | — | — | 75 | 21.6 | — | — | |
| BBTPBackbone=ResNet-1012022.10 | 58.9 | 30.4 | — | — | — | — | — | 75 | 21.6 | — | — | |
| FCRNLayer depth=50, Bootstrapping=true2016.05 | 58.7 | 42 | 53.8 | — | — | — | — | — | — | — | — | |
| BoxInstBackbone=ResNet-50, #Epoch=362023.10 | 58.6 | — | — | — | — | — | — | — | 34.6 | 34.3 | — | |
| AnnoCoInSupervision=Bounding box, Backbone=ResNet-101, Conditional Network=U-Net-based2020.07 | 58.2 | 34.3 | — | — | — | — | — | 73.8 | 32.1 | — | — | |
| AnnoCoIn (ResNet-101)Supervision=Bounding box, Backbone=ResNet-101, Conditional Network=ResNet-based2020.07 | 57.7 | 33.5 | — | — | — | — | — | 73.1 | 31.2 | — | — | |
| Arun et al.Supervision=Weak supervision: Bounding boxes2021.03 | 57.7 | 33.5 | — | — | — | — | — | 73.1 | 31.2 | — | — | |
| Arun et al.Backbone=ResNet-101, Arch=Mask R-CNN2021.05 | 57.7 | 33.5 | — | — | — | — | — | 73.1 | 31.2 | — | — | |
| Arun et al.Backbone=ResNet-1012022.10 | 57.7 | 33.5 | — | — | — | — | — | 73.1 | 31.2 | — | — | |
| FCRNLayer depth=502016.05 | 57.2 | 40.5 | 52.5 | — | — | — | — | — | — | — | — | |
| Sun et al.Supervision=Weak supervision: Bounding boxes2021.03 | 56.9 | — | — | — | — | — | 56.9 | — | 21.4 | — | — | |
| BESTIESupervision source=Point (P), MRCNN refinement=true2021.09 | 56.1 | 36.5 | — | — | — | — | — | 66.4 | 30.2 | — | — | |
| CIM (Ours)†Backbone=ResNet-50, Supervision=Image-level (I), Refinement=true2024.02 | 55.9 | 37.1 | — | — | — | — | — | 68.7 | 30.9 | — | — | |
| BBTPbackbone=ResNet-1012020.12 | 54.1 | — | — | — | — | — | — | — | 17.1 | 23.1 | — | |
| BBTPbackbone=ResNet-1012023.03 | 54.1 | — | — | — | — | — | — | — | 17.1 | 23.1 | — | |
| BBTPBackbone=ResNet-101, #Epoch=122023.10 | 54.1 | — | — | — | — | — | — | — | 17.1 | 23.1 | — | |
| CFMBackbone=ZF, Proposals=MCG2014.12 | 53.2 | — | — | — | — | — | — | — | — | — | — | |
| CIM (Ours)Backbone=HRNet-W48, Supervision=Image-level (I), Refinement=false2024.02 | 52.6 | 33.7 | — | — | — | — | — | 68.3 | 28.4 | — | — | |
| Mask R-CNNSupervision=pixel-level2019.07 | 51.4 | — | — | — | — | — | — | 58.9 | 32.4 | — | — | |
| Liao et al.Supervision=Weak supervision: Bounding boxes2021.03 | 51.3 | — | — | — | — | — | 51.9 | — | 22.4 | — | — | |
| CIM (Ours)Backbone=ResNet-50, Supervision=Image-level (I), Refinement=false2024.02 | 51.1 | 32.4 | — | — | — | — | — | 64.9 | 26.1 | — | — | |
| CFMBackbone=ZF, Proposals=SS2014.12 | 51 | — | — | — | — | — | — | — | — | — | — | |
| BESTIESupervision source=Image-level (I), MRCNN refinement=true2021.09 | 51 | 31.9 | — | — | — | — | — | 61.2 | 26.6 | — | — | |
| BESTIE†Backbone=HRNet-W48, Supervision=Image-level (I), Refinement=true2024.02 | 51 | 31.9 | — | — | — | — | — | 61.2 | 26.6 | — | — | |
| AnnoCoInSupervision=Image-level, Backbone=ResNet-50, Conditional Network=U-Net-based2020.07 | 50.9 | 30.2 | — | — | — | — | — | 59.7 | 28.5 | — | — | |
| Arun et al.Supervision source=Image-level (I), Proposal technique=Segment proposal (M)2021.09 | 50.9 | 30.2 | — | — | — | — | — | 59.7 | 28.5 | — | — | |
| Arun et al.†Backbone=ResNet-50, Supervision=Image-level (I), Refinement=true2024.02 | 50.9 | 30.2 | — | — | — | — | — | 59.7 | 28.5 | — | — | |
| CIM (Ours)Backbone=VGG-16, Supervision=Image-level (I), Refinement=false2024.02 | 50.8 | 31 | — | — | — | — | — | 65.6 | 25.2 | — | — | |
| SDS (AlexNet)backbone=AlexNet, proposal=MCG, time/img (s)=482015.12 | 49.7 | 25.3 | — | — | — | — | — | — | — | — | — | |
| SDS2016.05 | 49.7 | 25.3 | 41.4 | — | — | — | — | — | — | — | — | |
| SDSBackbone=AlexNet, Proposals=MCG2014.12 | 49.7 | — | — | — | — | — | — | — | — | — | — | |
| SDS2016.11 | 49.7 | 25.3 | — | — | — | — | — | — | — | — | — | |
| AnnoCoIn (ResNet-50)Supervision=Image-level, Backbone=ResNet-50, Conditional Network=ResNet-based2020.07 | 49.7 | 29.2 | — | — | — | — | — | 59.1 | 27.1 | — | — |