Multi-label classification on MS-COCO (val)
92.2mAPPanCAN
Evaluation results
| Method | Links | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| PanCANBackbone Network=CvT-W24 (22k), Resolution=384×3842025.12 | 92.2 | 89.4 | 83.9 | 86.6 | 89.7 | 85 | 87.5 | 82.3 | 82.3 | 94.8 | 72.6 | 95.4 | 72.5 | — | |
| CvT-W24(22k)Backbone Network=CvT-W24(22k), Resolution=384×3842025.12 | 91.3 | 88.8 | 83.2 | 85.9 | 89.2 | 84.6 | 86.8 | 81.6 | 81.6 | 93.9 | 72.1 | 94.7 | 72.1 | — | |
| CvT-W24Backbone Network=CvT-W24 (22k), Resolution=384×3842025.12 | 90.5 | 89.4 | 81.7 | 84.5 | 89.2 | 84.6 | 86.8 | 81.7 | 81.3 | 94.9 | 71.4 | 94.1 | 71.7 | — | |
| Q2L-SwinLBackbone Network=Swin-L (22k), Resolution=384×3842025.12 | 90.5 | 89.4 | 81.7 | 84.5 | 89.2 | 84.6 | 86.8 | 80.5 | 81.2 | 93.9 | 70.5 | 94.8 | 71 | — | |
| PanCANBackbone Network=TResNet-L, Resolution=448×4482025.12 | 90.1 | 87.2 | 82.3 | 84.6 | 87.2 | 84.9 | 84.7 | 79.5 | 80 | 92.5 | 69.8 | 93.5 | 71 | — | |
| Swin-LBackbone Network=Swin-L (22k), Resolution=384×3842025.12 | 89.5 | 86.9 | 89.4 | 81.6 | 86.3 | 83.1 | 84.7 | 81.3 | 81.3 | 93.9 | 70.8 | 94.1 | 71.5 | — | |
| Q2L-TResLBackbone Network=TResNet-L (22k), Resolution=448×4482025.12 | 89.2 | 86.3 | 81.4 | 83.8 | 86.5 | 83.3 | 84.9 | 79 | 80.2 | 91.6 | 69.4 | 92.9 | 70.5 | — | |
| MlTr-lBackbone Network=MlTr-l (22k), Resolution=384×3842025.12 | 88.5 | 86 | 81.7 | 83.4 | 84.9 | 82.1 | 83.9 | 80 | 81.1 | 93.6 | 69.9 | 94.3 | 71.1 | — | |
| GKGNetResolution=576 x 576, Param(M)=34.7, FLOPs (G)=40.1, Pre-training=ImageNet-1K2023.08 | 87.7 | 87 | 78.5 | 82.5 | 87.6 | 81 | 84.2 | 77.6 | 79.3 | — | — | — | — | — | |
| Q2L-TResLBackbone Network=TResNet-L, Resolution=448×4482025.12 | 87.3 | 87.6 | 76.5 | 81.6 | 88.4 | 78.5 | 83.1 | 77 | 78.5 | 91.9 | 66.2 | 93.5 | 67.6 | — | |
| PanCANBackbone Network=ResNet101, Resolution=576×5762025.12 | 86.9 | 86.1 | 77.2 | 81.4 | 87.2 | 79.5 | 82.9 | 76.7 | 78.5 | 90.6 | 66.5 | 92.6 | 68.5 | — | |
| GKGNetResolution=448 x 448, Param(M)=34.0, FLOPs (G)=21.9, Pre-training=ImageNet-1K2023.08 | 86.7 | 86.4 | 77.1 | 81.5 | 87.3 | 79.7 | 83.3 | 77 | 78.8 | — | — | — | — | — | |
| ASLBackbone Network=TResNet-L, Resolution=448×4482025.12 | 86.6 | 87.2 | 76.4 | 81.4 | 88.2 | 79.2 | 81.8 | 75.1 | 77.4 | 91.8 | 63.4 | 92.9 | 66.4 | — | |
| Q2L-R101Resolution=576 x 576, Param(M)=193.6, FLOPs (G)=80.8, Pre-training=ImageNet-1K, EMA=true2023.08 | 86.5 | 85.8 | 76.7 | 81 | 87 | 78.9 | 82.8 | 76.5 | 78.3 | — | — | — | — | — | |
| Q2L-R101Backbone Network=ResNet101, Resolution=576×5762025.12 | 86.5 | 85.8 | 76.7 | 81 | 87 | 79.8 | 82.8 | 76.5 | 78.3 | 90.4 | 66.3 | 92.4 | 67.9 | — | |
| IDA-SwinS(H)Resolution=576 x 576, Param(M)=48.9, FLOPs (G)=64.0, Pre-training=ImageNet-1K2023.08 | 86.4 | 69.6 | 86.5 | 77.1 | 69.6 | 88.8 | 78.1 | 75.6 | 77.8 | — | — | — | — | — | |
| IDA-R101(H)Resolution=576 x 576, Param(M)=53.2, FLOPs (G)=54.1, Pre-training=ImageNet-1K2023.08 | 86.3 | — | — | — | — | — | 82.5 | 76.4 | 78.2 | — | — | — | — | — | |
| TDRGResolution=576 x 576, Param(M)=68.3, FLOPs (G)=69.8, Pre-training=ImageNet-1K2023.08 | 86 | 87 | 74.7 | 80.4 | 87.5 | 77.9 | 82.4 | 76.2 | 78.1 | — | — | — | — | — | |
| ADD-GCNBackbone Network=ResNet101, Resolution=576×5762025.12 | 85.7 | 84.8 | 74.5 | 79.6 | 86.6 | 76.9 | 82 | 75.8 | 77.9 | 86.6 | 65.6 | 90.6 | 70.5 | — | |
| IDA-SwinS(H)Resolution=448 x 448, Param(M)=48.9, FLOPs (G)=35.0, Pre-training=ImageNet-1K2023.08 | 85.5 | 68.9 | 85.6 | 76.3 | 69.8 | 87.5 | 77.7 | 75.2 | 77.4 | — | — | — | — | — | |
| ADD-GCNResolution=576 x 576, Param(M)=48.2, FLOPs (G)=52.7, Pre-training=ImageNet-1K2023.08 | 85.2 | 84.7 | 75.9 | 80.1 | 84.9 | 79.4 | 82 | 75.8 | 77.9 | — | — | — | — | — | |
| C-TranResolution=576 x 576, Param(M)=120.4, FLOPs (G)=84.2, Pre-training=ImageNet-1K2023.08 | 85.1 | 86.3 | 74.3 | 79.9 | 87.7 | 76.5 | 81.7 | 76 | 77.6 | — | — | — | — | — | |
| C-TransBackbone Network=ResNet101, Resolution=576×5762025.12 | 85.1 | 84.3 | 74.9 | 79.7 | 87.7 | 76.5 | 81.7 | 76 | 77.6 | 90.1 | 65.7 | 92.1 | 71.4 | — | |
| Q2L-R101Resolution=448 x 448, Param(M)=193.6, FLOPs (G)=51.4, Pre-training=ImageNet-1K, EMA=true2023.08 | 84.9 | 84.8 | 74.5 | 79.3 | 86.6 | 76.9 | 81.5 | 73.3 | 75.4 | — | — | — | — | — | |
| Q2L-R101Backbone Network=ResNet101, Resolution=448×4482025.12 | 84.9 | 84.8 | 74.5 | 79.3 | 86.6 | 76.9 | 81.5 | 73.3 | 75.4 | 78 | 69.1 | 80.7 | 70.8 | — | |
| TDRGResolution=448 x 448, Param(M)=68.3, FLOPs (G)=42.2, Pre-training=ImageNet-1K2023.08 | 84.6 | 86 | 73.1 | 79 | 86.6 | 76.4 | 81.2 | 75 | 77.2 | — | — | — | — | — | |
| MCARResolution=576 x 576, Pre-training=ImageNet-1K2023.08 | 84.5 | 84.3 | 73.9 | 78.7 | 86.9 | 76.1 | 81.1 | 75.3 | 77 | — | — | — | — | — | |
| KGGRResolution=576 x 576, Pre-training=ImageNet-1K2023.08 | 84.3 | 85.6 | 72.7 | 78.6 | 87.1 | 75.6 | 80.9 | 75 | 77 | — | — | — | — | — | |
| MS-CMAResolution=448 x 448, Pre-training=ImageNet-1K2023.08 | 83.8 | 82.9 | 74.4 | 78.4 | 84.4 | 77.9 | 81 | 74.9 | 77.1 | — | — | — | — | — | |
| MCARResolution=448 x 448, Pre-training=ImageNet-1K2023.08 | 83.8 | 85 | 72.1 | 78 | 88 | 73.9 | 80.3 | 75.1 | 76.7 | — | — | — | — | — | |
| SSGRLResolution=576 x 576, Param(M)=92.3, FLOPs (G)=68.5, Pre-training=ImageNet-1K2023.08 | 83.8 | 91.9 | 62.5 | 72.7 | 93.8 | 64.1 | 76.2 | 76.8 | 79.7 | — | — | — | — | — | |
| MS-CMABackbone Network=ResNet101, Resolution=448×4482025.12 | 83.8 | 84.9 | 74.4 | 78.4 | 84.4 | 77.9 | 81 | 74.3 | 77.2 | 86.7 | 64.9 | 90.9 | 67.2 | — | |
| MCARBackbone Network=ResNet101, Resolution=448×4482025.12 | 83.8 | 84.1 | 74.8 | 79.3 | 87.1 | 78 | 83.1 | 75.9 | 76.7 | 88.1 | 65.7 | 90.6 | 66.3 | — | |
| SSGRLBackbone Network=ResNet101, Resolution=576×5762025.12 | 83.8 | 85.9 | 74.8 | 80 | 87.7 | 79.1 | 83.7 | 72.7 | 76.2 | 91.9 | 62.5 | 93.8 | 64.1 | — | |
| KSSNetResolution=448 x 448, Param(M)=173.8, Pre-training=ImageNet-1K2023.08 | 83.7 | 84.6 | 73.2 | 77.2 | 87.8 | 76.2 | 81.5 | — | — | — | — | — | — | — | |
| KSSNetBackbone Network=ResNet101, Resolution=448×4482025.12 | 83.7 | 84.1 | 73.1 | 78.2 | 87.8 | 76.2 | 81.5 | — | — | — | — | — | — | — | |
| ML-GCNResolution=448 x 448, Param(M)=44.9, FLOPs (G)=31.5, Pre-training=ImageNet-1K2023.08 | 83 | 85.1 | 72 | 78 | 85.8 | 75.4 | 80.3 | 74.6 | 76.7 | — | — | — | — | — | |
| ML-GCNBackbone Network=ResNet101, Resolution=448×4482025.12 | 83 | 85.1 | 72.2 | 78 | 85.1 | 74.7 | 79.4 | 74.2 | 76.3 | 86.3 | 63.4 | 89.1 | 66.7 | — | |
| KLUE (v1)Backbone=SWIN, LSAT=false, External Rules=false2026.04 | 82.78 | — | — | — | — | — | — | — | — | — | — | — | — | 98.3 | |
| KLUE (v1)Backbone=SWIN, LSAT=true, External Rules=false2026.04 | 82.71 | — | — | — | — | — | — | — | — | — | — | — | — | 98.23 | |
| CADMResolution=448 x 448, Pre-training=ImageNet-1K2023.08 | 82.3 | 82.5 | 72.2 | 77 | 84 | 75.6 | 79.6 | 73.5 | 76 | — | — | — | — | — | |
| CADMBackbone Network=ResNet101, Resolution=448×4482025.12 | 82.3 | 85.2 | 72 | 78 | 85.6 | 74.6 | 79.6 | 73.5 | 76 | 87.1 | 63.6 | 89.4 | 66 | — | |
| BaselineBackbone=SWIN, LSAT=false, External Rules=false2026.04 | 82.24 | — | — | — | — | — | — | — | — | — | — | — | — | 98.29 | |
| KLUE (v2)Backbone=SWIN, LSAT=true, External Rules=false2026.04 | 82.22 | — | — | — | — | — | — | — | — | — | — | — | — | 98.13 | |
| GKGNetResolution=224 x 224, Param(M)=33.3, FLOPs (G)=5.2, Pre-training=ImageNet-1K2023.08 | 82 | 81.7 | 73.1 | 77.1 | 83.1 | 76.3 | 79.6 | 73.7 | 76.1 | — | — | — | — | — | |
| KLUE (v2)Backbone=SWIN, LSAT=false, External Rules=false2026.04 | 81.93 | — | — | — | — | — | — | — | — | — | — | — | — | 98.13 | |
| MltrResolution=224 x 224, Param(M)=33.0, Pre-training=ImageNet-1K2023.08 | 81.9 | 80.7 | 71.5 | 75.2 | 81.4 | 76.3 | 78.1 | — | — | — | — | — | — | — | |
| KLUE (v2)Backbone=WRN-101, LSAT=true, External Rules=false2026.04 | 80.8 | — | — | — | — | — | — | — | — | — | — | — | — | 97.91 | |
| KLUE (v1)Backbone=WRN-101, LSAT=true, External Rules=false2026.04 | 80.79 | — | — | — | — | — | — | — | — | — | — | — | — | 97.88 | |
| IDA-SwinS(H)Resolution=224 x 224, Param(M)=48.9, FLOPs (G)=8.7, Pre-training=ImageNet-1K2023.08 | 80.6 | 64.5 | 81.1 | 71.8 | 65.8 | 83.9 | 73.8 | 71.4 | 74.5 | — | — | — | — | — | |
| LTN (v1)Backbone=SWIN, LSAT=false, External Rules=true2026.04 | 80.47 | — | — | — | — | — | — | — | — | — | — | — | — | 98.03 | |
| LTN (v2)Backbone=SWIN, LSAT=false, External Rules=true2026.04 | 79.46 | — | — | — | — | — | — | — | — | — | — | — | — | 97.64 | |
| ResNet-101Resolution=224 x 224, Param(M)=44.5, FLOPs (G)=7.8, Pre-training=ImageNet-1K2023.08 | 78.3 | 80.2 | 66.7 | 72.8 | 83.9 | 70.8 | 76.8 | 69.7 | 73.6 | — | — | — | — | — | |
| ResNet-101Backbone Network=ResNet101, Resolution=224×2242025.12 | 78.3 | 80.2 | 66.7 | 72.8 | 83.9 | 70.8 | 76.6 | 69.7 | 73.6 | 84.1 | 59.4 | 89.1 | 62.8 | — | |
| SRNResolution=224 x 224, Param(M)=76.8, FLOPs (G)=9.0, Pre-training=ImageNet-1K2023.08 | 77.1 | 81.6 | 65.4 | 71.2 | 82.7 | 69.9 | 75.8 | 67.4 | 72.9 | — | — | — | — | — | |
| SRNBackbone Network=ResNet101, Resolution=224×2242025.12 | 77.1 | 81.6 | 65.4 | 71.2 | 82.7 | 69.9 | 75.8 | 67.4 | 72.9 | 85.2 | 58.8 | 87.4 | 62.5 | — | |
| KLUE (v1)Backbone=WRN-101, LSAT=false, External Rules=false2026.04 | 74.06 | — | — | — | — | — | — | — | — | — | — | — | — | 97.08 | |
| KLUE (v2)Backbone=WRN-101, LSAT=false, External Rules=false2026.04 | 72.96 | — | — | — | — | — | — | — | — | — | — | — | — | 97.05 | |
| BaselineBackbone=WRN-101, LSAT=false, External Rules=false2026.04 | 71.97 | — | — | — | — | — | — | — | — | — | — | — | — | 96.83 | |
| LTN (v1)Backbone=WRN-101, LSAT=false, External Rules=true2026.04 | 41.08 | — | — | — | — | — | — | — | — | — | — | — | — | 92.46 | |
| LTN (v2)Backbone=WRN-101, LSAT=false, External Rules=true2026.04 | 40.84 | — | — | — | — | — | — | — | — | — | — | — | — | 92.4 |