PP-YOLOE: An evolved version of YOLO

Benchmark Model Rank Results
multi-object-tracking-on-mot16PPTracking#1MOTA: 77.7
multiple-object-tracking-on-crohdPP-Tracking#1MOTA: 72.6
online-multi-object-tracking-on-mot16PP-Tracking#1MOTA: 77.7
real-time-object-detection-on-coco-test-devPP-YOLOE+-X (Objects365 pre-training)#8box AP: 54.7FPS (batch 1, GPU in brackets): 45 (V100)
real-time-object-detection-on-coco-test-devPP-YOLOE+-L (Objects365 pre-training)#13box AP: 52.9FPS (batch 1, GPU in brackets): 78 (V100)
real-time-object-detection-on-coco-test-devPP-YOLOE+-M (Objects365 pre-training)#18box AP: 49.8FPS (batch 1, GPU in brackets): 123 (V100)
object-detection-on-coco-test-devPP-YOLOE-x(CSPRepResNet-x, 640x640, single-scale )#69box mAP: 52.2AP50: 69.9AP75: 56.5APS: 33.3APM: 56.3APL: 66.4
object-detection-on-coco-test-devPP-YOLOE-l(CSPRepResNet-l, 640x640, single-scale )#74box mAP: 51.4AP50: 68.9AP75: 55.6APS: 31.4APM: 55.3APL: 66.1
object-detection-on-coco-test-devPP-YOLOE-m(CSPRepResNet-m, 640x640, single-scale )#94box mAP: 48.9AP50: 66.5AP75: 53.0APS: 28.6APM: 52.9APL: 63.8
object-detection-on-coco-test-devPP-YOLOE-s(CSPRepResNet-s, 640x640, single-scale )#146box mAP: 43.1AP50: 60.5AP75: 46.6APS: 23.2APM: 46.4APL: 56.9