Research on Camouflaged Human Target Detection Based on Deep Learning

Table 4

Detection performance of different models.

Module	mAP/%	FPS	Size/MB
RetinaNet	81.4	6.4	138
YOLOX	88.7	13.4	34.3
YOLOv5s	90.3	125	14.4
YOLOv5m	92.6	45.5	40.4
YOLOv5l	93.4	43.5	93.7
MC-YOLOv5s	94	93.8	33.1