RepPoints: Point Set Representation for Object Detection

Benchmark Model Rank Results
object-detection-on-coco-test-devRPDet (ResNet-101-DCN, multi-scale)#115box mAP: 46.5AP50: 67.4AP75: 50.9APS: 30.3APM: 49.7APL: 57.1
object-detection-on-coco-test-devRPDet (ResNet-101-DCN)#150box mAP: 42.8AP50: 65.0AP75: 46.3APS: 24.9APM: 46.2APL: 54.7
object-detection-on-coco-test-devRPDet (ResNet-101)#166box mAP: 41AP50: 62.9AP75: 44.3APS: 23.6APM: 44.1APL: 51.7
object-detection-on-coco-val2017RPDet (ResNeXt-101-DCN, multi-scale)#283box AP: 46.8
object-detection-on-coco-val2017RPDet (ResNet-101-DCN, multi-scale)#288box AP: 46.4
object-detection-on-coco-val2017RPDet (ResNet-101-DCN, multi-scale train)#318box AP: 44.8
object-detection-on-coco-val2017RPDet (ResNeXt-101-DCN)#326box AP: 44.5
object-detection-on-coco-val2017RPDet (ResNet-50, multi-scale train)#387box AP: 40.8
object-detection-on-coco-val2017RPDet (ResNet-101)#399box AP: 40.3
object-detection-on-coco-val2017RPDet (ResNet-50)#416box AP: 38.6