List your company

Panoptic Segmentation on COCO 2017 (val)

59.9PQMask2Former
42.7447.19551.6556.105Jan 2, 2020Nov 15, 2020Sep 29, 2021Aug 13, 2022Jun 27, 2023May 10, 2024Mar 24, 2025

Evaluation results

MethodLinks
Mask2FormerBackbone=ViT-Adapter-g+, Pre-training=DINOv2, Params=1216M, Input size=1280x12802025.0359.9————————————————————————13,7906—
Mask2Former+Backbone=ViT-Adapter-L+, Pre-training=DINOv2, Params=354M, Input size=1280x12802025.0359.7————————————————————————4,81710—
Mask DINOBackbone=Swin-L [38], #Params=223M, Extra Data=true, Training Protocol=Individual Training2022.1159.4———————————————————————————
EoMTBackbone=ViT-g, Pre-training=DINOv2, Params=1171M, Input size=1280x12802025.0359.2————————————————————————12,71212—
MaskDINOBackbone=Swin-L, Pre-training=IN21K, Params=223M, Input size=800x8002025.0358.3————————————————————————1,32614—
EoMTBackbone=ViT-L, Pre-training=DINOv2, Params=322M, Input size=1280x12802025.0358.3————————————————————————4,14630—
VAN-B6Backbone=VAN-B6, Base Architecture=Mask2Former [113], Query type=200 queries, Epochs=50, Pre-trained=ImageNet-22K2022.0258.2——64.8——48.2—————————————————————
kMaX-DeepLabBackbone=ConvNeXt-L [39], #Params=232M, Extra Data=true, Training Protocol=Individual Training2022.1158.1——64.3——48.8—————————————————————
MP-Formerbackbone=Swin-L, query type=200 queries, epochs=72, #params.=216M, FLOPs=868G, fps=42023.0358.1——64.4——48.4——————————67.648.9—————————
kMaX-DeepLabBackbone=ConvNeXt-L [39], #Params=232M, Extra Data=false, Training Protocol=Individual Training2022.1158——64.2——48.6—————————————————————
OneFormerBackbone=DINAT-L [21], #Params=223M, Extra Data=false, Training Protocol=Joint Training2022.1158——64.3——48.4—————————————————————
kMaX-DeepLab*Backbone=ConvNeXt-L, #Params=232M, #FLOPs=749G, #Queries=256, Epochs=812022.1158——64.2——48.6—————————————————————
OneFormerBackbone=DINAT-L+, #Params=223M, #FLOPs=736G, #Queries=150, Epochs=1002022.1158——64.3——48.4—————————————————————
kMaX-DeepLabBackbone=ConvNext-L, Pre-training=IN21K, Params=232M, Input size=1281x12812025.0358———————————————————————————
OneFormer+Backbone=DINAT-L, Pre-training=IN21K, Params=223M, Input size=800x8002025.0358————————————————————————73620—
OneFormerBackbone=Swin-L [38], #Params=219M, Extra Data=false, Training Protocol=Joint Training2022.1157.9——64.4——48—————————————————————
kMaX-DeepLab*Backbone=ConvNeXt-L, #Params=232M, #FLOPs=749G, #Queries=128, Epochs=812022.1157.9——64——48.6—————————————————————
OneFormerBackbone=Swin-L+, #Params=219M, #FLOPs=891G, #Queries=150, Epochs=1002022.1157.9——64.4——48—————————————————————
Mask2Formerbackbone=Swin-L*, query type=100 queries, epochs=100, #params=216M, FLOPS=868G, fps=42021.1257.8——64.2——48.1—————————48.667.4——————————
Swin-LBackbone=Swin-L, Base Architecture=Mask2Former [113], Query type=200 queries, Epochs=50, Pre-trained=ImageNet-22K2022.0257.8——64.2——48.1—————————————————————
Mask2Former2022.0557.8———————————————————————————
Mask2Former-PanopticBackbone=Swin-L [38], #Params=216M, Extra Data=false, Training Protocol=Individual Training2022.1157.8——64.2——48.1—————————————————————
Mask2Former-PanopticBackbone=Swin-L, #Params=216M, #FLOPs=875G, #Queries=200, Epochs=1002022.1157.8——64.2——48.1—————————————————————
Mask2Formerbackbone=Swin-L, query type=200 queries, epochs=100, #params.=216M, FLOPs=868G, fps=42023.0357.8——64.2——48.1——————————67.448.6—————————
Mask2Formermodel category=specialist model, venue=CVPR'222023.0457.8———————————————————————————
Mask2FormerBackbone=Swin-L, Pre-training=IN21K, Params=216M, Input size=800x8002025.0357.8————————————————————————86824—
Mask2FormerBackbone=ViT-Adapter-g+, Pre-training=DINOv2, Params=1209M, Input size=640x6402025.0357.7————————————————————————2,51020—
Mask2Former+Backbone=ViT-Adapter-L+, Pre-training=DINOv2, Params=349M, Input size=640x6402025.0357.1————————————————————————83029—
EoMTBackbone=ViT-g, Pre-training=DINOv2, Params=1164M, Input size=640x6402025.0357————————————————————————2,26155—
SegMAN-S EncoderEncoder=SegMAN-S Encoder, Parameters=48.3M, GFLOPs=2832024.1256.8———————————————————————————
EoMTBackbone=ViT-L, Pre-training=DINOv2, Params=316M, Input size=640x6402025.0356————————————————————————669128—
Panoptic SegFormerBackbone=Swin-L [38], #Params=221M, Extra Data=false, Training Protocol=Individual Training2022.1155.8——61.7——46.9—————————————————————
Panoptic SegFormerBackbone=Swin-L, #Params=221M, #FLOPs=816G, #Queries=353, Epochs=242022.1155.8——61.7——46.9—————————————————————
CMT-DeepLab (iter 200k)backbone=Axial-R104-RFN, TTA=false, params=270.3M2022.0655.3——61——46.6—————————————————————
CMT-DeepLabbackbone=Axial-R104-RFN, TTA=false, params=270.3M2022.0655.1——60.6——46.8—————————————————————
VAN-B2Backbone=VAN-B2, Base Architecture=Mask2Former [113], Query type=100 queries, Epochs=50, Pre-trained=ImageNet-1K2022.0254.9——61.2——45.3—————————————————————
MiT-B2Encoder=MiT-B2, Parameters=48.5M, GFLOPs=3152024.1254.7———————————————————————————
K-NetBackbone=Swin-L, Box-free=true, NMS-free=true, Epochs=362021.0654.6——60.2——46—————————————————————
K-Netbackbone=Swin-L+, query type=100 queries, epochs=362021.1254.6——60.2——46—————————————————————
K-NetBackbone=Swin-L+, #Queries=100, Epochs=362022.1154.6——60.2——46—————————————————————
K-Netbackbone=Swin-L, query type=100 queries, epochs=362023.0354.6——60.2——46—————————————————————
CMT-DeepLabbackbone=Axial-R104, TTA=false, params=135.2M2022.0654.1——58.8——47.1—————————————————————
ReMaxBackbone=Res502024.0453.5———————————————————————————
Swin-TBackbone=Swin-T, Base Architecture=Mask2Former [113], Query type=100 queries, Epochs=50, Pre-trained=ImageNet-1K2022.0253.2——59.3——44—————————————————————
CleanBackbone=Swin-T2024.1053.2———————————————43.363.2——————————
CMT-DeepLabbackbone=Axial-R50 (ImageNet-22K), TTA=false, params=94.9M2022.0653——57.7——45.9—————————————————————
ResNet50Encoder=ResNet50, Parameters=52M, GFLOPs=2862024.1253———————————————————————————
MP-Formerbackbone=R101, query type=100 queries, epochs=50, #params.=63M, FLOPs=293G, fps=7.22023.0352.9——58.7——44——————————61.843—————————
Lite-ECO-M2FBackbone=SWIN-T, Total GFLOPs=178.43, Encoder GFLOPs=64.422024.0452.84———————————————————————————
MaskFormerbackbone=Swin-L*, query type=100 queries, epochs=300, #params=212M, FLOPS=792G, fps=5.22021.1252.7——58.5——44—————————40.164.8——————————
MaskFormerbackbone=Swin-L (ImageNet-22K), TTA=false, params=212M2022.0652.7——58.5——44—————————————————————
MaskFormerBackbone=Swin-L, #Params=212M, #FLOPs=792G, #Queries=100, Epochs=3002022.1152.7——58.5——44—————————————————————
MaskFormerbackbone=Swin-L, query type=100 queries, epochs=300, #params.=212M, FLOPs=792G, fps=5.22023.0352.7——58.5——44——————————64.840.1—————————
Lite-M2FBackbone=SWIN-T, Total GFLOPs=188.00, Encoder GFLOPs=79.782024.0452.7———————————————————————————
Mask2Formerbackbone=R101, query type=100 queries, epochs=50, #params=63M, FLOPS=293G, fps=7.22021.1252.6——58.5——43.7—————————42.662.4——————————
Mask2Formerbackbone=R101, query type=100 queries, epochs=50, #params.=63M, FLOPs=293G, fps=7.22023.0352.6——58.5——43.7——————————62.442.6—————————
MP-Formerbackbone=R50, query type=100 queries, epochs=50, #params.=44M, FLOPs=226G, fps=8.62023.0352.2——58.3——42.9——————————61.642.5—————————
ECO-M2F (β = 0.0005)Backbone=SWIN-T, Total GFLOPs=202.39, Encoder GFLOPs=88.472024.0452.06———————————————————————————
M2FBackbone=SWIN-T, Total GFLOPs=235.57, Encoder GFLOPs=121.692024.0452.03———————————————————————————
Mask2Formerbackbone=R50, query type=100 queries, epochs=50, #params=44M, FLOPS=226G, fps=8.62021.1251.9——57.7——43—————————41.761.7——————————
CleanBackbone=ResNet502024.1051.9———————————————41.761.7——————————
ECO-M2FBackbone=Res50, Total GFLOPs=195.55, Encoder GFLOPs=92.372024.0451.89———————————————————————————
MaskFormerbackbone=Swin-B (ImageNet-22K), TTA=false, params=102M2022.0651.8——56.9——44.1—————————————————————
M2FBackbone=Res50, Total GFLOPs=229.10, Encoder GFLOPs=135.002024.0451.73———————————————————————————
Mask2Formerbackbone=R50, query type=100 queries, epochs=50, #params.=44M, FLOPs=226G, fps=8.62023.0351.4——57.2——42.7——————————61.541.9—————————
Mask DINO w/ Mr. DETR++Backbone=ResNet-50, Epochs=12, Queries=3002024.1251.2——57.5——41.7————————————————————46
Max-DeepLabbackbone=Max-L, query type=128 queries, epochs=216, #params=451M, FLOPS=3692G2021.1251.1——57——42.2—————————————————————
MaX-DeepLab-Lbackbone=MaX-L, TTA=false, params=451M2022.0651.1——57——42.2—————————————————————
Max-DeepLabbackbone=Max-L, query type=128 queries, epochs=216, #params.=451M, FLOPs=3692G2023.0351.1——57——42.2—————————————————————
ECO-M2F (β = 0.02)Backbone=SWIN-T, Total GFLOPs=181.64, Encoder GFLOPs=67.712024.0450.79———————————————————————————
Mask DINO w/ Mr. DETRBackbone=ResNet-50, Epochs=12, Queries=3002024.1250.2——56.6——40.4————————————————————45.4
K-NetBackbone=R101-FPN, Box-free=true, NMS-free=true, Epochs=362021.0649.6——55.1——41.4—————————————————————
Mask DINOBackbone=ResNet-50, Epochs=12, Queries=3002024.1249——54.8——40.2————————————————————43.2
MaX-DeepLab-Sbackbone=MaX-S, TTA=false, params=61.9M2022.0648.4——53——41.5—————————————————————
YOSOBackbone=Res50, Total GFLOPs=114.502024.0448.4———————————————————————————
K-NetBackbone=R101-FPN-DCN, Box-free=true, NMS-free=true, Epochs=362021.0648.3——54——39.7—————————————————————
Panoptic SegformerBackbone=ResNet-50, Epochs=12, Queries=3532024.1248——52.3——41.5—————————————————————
MaskFormerbackbone=R101, query type=100 queries, epochs=300, #params=64M, FLOPS=248G, fps=142021.1247.6——52.5——40.3—————————34.159.3——————————
MaskFormerbackbone=R101, query type=100 queries, epochs=300, #params.=64M, FLOPs=248G, fps=142023.0347.6——52.5——40.3——————————59.334.1—————————
K-NetBackbone=R50-FPN, Box-free=true, NMS-free=true, Epochs=362021.0647.1——51.7——40.3—————————————————————
Mask2FormerBackbone=ResNet-50, Epochs=12, Queries=1002024.1246.9——52.5——38.4—————————————————————
RAP-SAMBackbone=Res50, Total GFLOPs=123.002024.0446.9———————————————————————————
MaskFormerbackbone=R50, query type=100 queries, epochs=300, #params=45M, FLOPS=181G, fps=17.62021.1246.5——51——39.8—————————3357.8——————————
MaskFormerbackbone=R50, query type=100 queries, epochs=300, #params.=45M, FLOPs=181G, fps=17.62023.0346.5——51——39.8——————————57.833—————————
MFBackbone=Res50, Total GFLOPs=181.002024.0446.5———————————————————————————
PEMBackbone=Res50, Total GFLOPs=110.902024.0446.38———————————————————————————
UViMBackbone=ViT-L/16 (Encoder), ViT-B (Decoder), Pre-training=ImageNet-21k, Input Resolution=1280x12802022.0545.8———————————————————————————
UVIMmodel category=specialist model, venue=NeurIPS'222023.0445.8———————————————————————————
DETR-R101Backbone=R1012020.0545.179.955.550.580.961.73778.54633——————————————————
DETRbackbone=R101, query type=100 queries, epochs=500+252021.1245.1——50.5——37—————————33———————————
DETRbackbone=R101, TTA=false, params=61.8M2022.0645.1——50.5——37—————————————————————
DETR-R101Backbone=ResNet-1012022.0545.1———————————————————————————
DETRbackbone=R101, query type=100 queries, epochs=500+252023.0345.1——50.5——37———————————33—————————
DETR-DC5Backbone=R50, Dilated C5 (DC5)=true2020.0544.679.85549.480.560.637.378.746.531.9——————————————————
BlendMaskBackbone=R-1012020.0144.380.153.451.6——33.2—————————————————————
PanopticFPN++Backbone=R101, Schedule=18x2020.0544.179.553.35183.260.633.67442.139.7——————————————————
Axial-DeepLab-Lbackbone=AX-L, TTA=true, params=44.9M2022.0643.9——48.6——36.8—————————————————————
Panoptic FCNBackbone=R50-FPN, Box-free=true, NMS-free=true, Epochs=362021.0643.6——49.3——35—————————————————————
DETRBackbone=R502020.0543.479.353.848.279.859.536.378.545.331.1——————————————————
DETRBackbone=R50, Box-free=true, NMS-free=true, Epochs=300+252021.0643.4——48.2——36.3—————————————————————