Panoptic Segmentation on MS-COCO 2018 (test-dev)
53.2PQ (Panoptic Quality)Megvii (Face++)
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Megvii (Face++)backbone=ensemble model, extra data=true, e2e=false, enhanceTh=true, enhancest=true2018.12 | 53.2 | 83.2 | 62.9 | 62.2 | 85.5 | 72.5 | 39.5 | 79.7 | 48.5 | |
| Megviibackbone=ensemble2019.11 | 53.2 | 83.2 | 62.9 | — | — | — | — | — | — | |
| SOGNetbackbone=ResNet-101-FPN, extra deformable convolution layers=true, longer training schedule=true2019.11 | 47.8 | 80.7 | 57.6 | — | — | — | — | — | — | |
| Caribbeanbackbone=ensemble model, extra data=false, e2e=false, enhanceTh=true, enhancest=true2018.12 | 46.8 | 80.5 | 57.1 | 54.3 | 81.8 | 65.9 | 35.5 | 78.5 | 43.8 | |
| Caribbeanbackbone=ensemble2019.11 | 46.8 | 80.5 | 57.1 | — | — | — | — | — | — | |
| UPSNetbackbone=ResNet-101-FPN, extra deformable convolution layers=true, longer training schedule=true2019.11 | 46.6 | 80.5 | 56.9 | — | — | — | — | — | — | |
| AUNetbackbone=ResNeXt-152-FPN, extra data=false, e2e=true, enhanceTh=false, enhancest=false2018.12 | 46.5 | 81 | 56.1 | 55.8 | 83.7 | 66.3 | 32.5 | 77 | 40.7 | |
| AUNetbackbone=ResNeXt-152-FPN2019.11 | 46.5 | 81 | 56.1 | — | — | — | — | — | — | |
| PKU_360backbone=ResNeXt-152-FPN, extra data=false, e2e=false, enhanceTh=true, enhancest=true2018.12 | 46.3 | 79.6 | 56.1 | 58.6 | 83.7 | 69.6 | 27.6 | 73.6 | 35.6 | |
| PKU-360backbone=ResNeXt-152-FPN2019.11 | 46.3 | 79.6 | 56.1 | — | — | — | — | — | — | |
| AUNetbackbone=ResNet-152-FPN, extra data=false, e2e=true, enhanceTh=false, enhancest=false2018.12 | 45.5 | 80.8 | 55 | 54.7 | 83.4 | 65.2 | 31.6 | 76.9 | 39.7 | |
| AUNetbackbone=ResNet-101-FPN, extra data=false, e2e=true, enhanceTh=false, enhancest=false2018.12 | 45.2 | 80.6 | 54.7 | 54.4 | 83.3 | 64.8 | 31.3 | 76.6 | 39.4 | |
| OANetbackbone=ResNet-101-FPN2019.11 | 41.3 | — | — | — | — | — | — | — | — | |
| Panoptic FPNbackbone=ResNet-101-FPN2019.11 | 40.9 | — | — | — | — | — | — | — | — | |
| JSIS-Netbackbone=ResNet-50, extra data=false, e2e=true, enhanceTh=false, enhancest=false2018.12 | 27.2 | 71.9 | 35.9 | 29.6 | 71.6 | 39.4 | 23.4 | 72.3 | 30.6 |