Cascade R-CNN: Delving into High Quality Object Detection

Benchmark Model Rank Results
2d-object-detection-on-sardet-100kCascade R-CNN#5box mAP: 51.1
object-detection-on-ai-todCascade R-CNN (ResNet-50-FPN)#6AP: 13.8AP50: 30.8AP75: 10.5APvt: 0.0APt: 10.6APs: 25.5
object-detection-on-cocoCascade R-CNN (ResNet-101-FPN+, cascade)#162box mAP: 42.8AP50: 62.1AP75: 46.3APS: 23.7APM: 45.5APL: 55.2
object-detection-on-cocoCascade R-CNN (ResNet-50-FPN+, cascade)#185box mAP: 40.6AP50: 59.9AP75: 44APS: 22.6APM: 42.7APL: 52.1
object-detection-on-cocoCascade R-CNN (ResNet-101-FPN+)#205box mAP: 38.8AP50: 61.1AP75: 41.9APS: 21.3APM: 41.8APL: 49.8
object-detection-on-cocoCascade R-CNN (ResNet-50-FPN+)#217box mAP: 36.5AP50: 59AP75: 39.2APS: 20.3APM: 38.8APL: 46.4
object-detection-on-coco-minivalCascade R-CNN (ResNet-101-FPN+, cascade)#149box AP: 42.7AP50: 61.6AP75: 46.6APS: 23.8APM: 46.2APL: 57.4
object-detection-on-coco-minivalCascade R-CNN (ResNet-50-FPN+)#177box AP: 40.3AP50: 59.4AP75: 43.7APS: 22.9APM: 43.7APL: 54.1