Object Detection on COCO 2016 2017 (test-dev)
47.4AP (Box)PANet
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| PANetBackbone=ResNeXt-101, ms-train=true2018.03 | 47.4 | 67.2 | 51.8 | 30.1 | 51.7 | 60 | |
| PANetBackbone=ResNeXt-101, ms-train=false2018.03 | 45 | 65 | 48.6 | 25.4 | 48.6 | 59.1 | |
| PANetBackbone=ResNet-50, ms-train=true2018.03 | 42.5 | 62.3 | 46.4 | 26.3 | 47 | 52.3 | |
| Champion 2016Backbone=2xResNet-101 + 3xInception-ResNet-v22018.03 | 41.6 | 62.3 | 45.6 | 24 | 43.9 | 55.2 | |
| PANetBackbone=ResNet-50, ms-train=false2018.03 | 41.2 | 60.4 | 44.4 | 22.7 | 44 | 54.6 | |
| Mask R-CNN + FPNBackbone=ResNeXt-101, ms-train=false2018.03 | 39.8 | 62.3 | 43.4 | 22.1 | 43.2 | 51.2 | |
| RetinaNetBackbone=ResNet-101, ms-train=false2018.03 | 39.1 | 59.1 | 42.3 | 21.8 | 42.7 | 50.2 | |
| Mask R-CNN + FPNBackbone=ResNet-101, ms-train=false2018.03 | 38.2 | 60.3 | 41.7 | 20.1 | 41.1 | 50.2 |