Performance Evaluation of Deep Learning Algorithm Using High-End Media Processing Board in Real-Time Environment

Table 4

Comparison of various convolutional neural network (CNN) frameworks in terms of map, FPS, and layers.

Convolutional neural network (CNN) frameworks Comparision
CNN variants	Author	mAP(mean average precision)	FPS(frame per second)	Layers
YOLO	Redmon et al. [34]	63.4	45	26
YOLOv2	Redmon et al. [42]	48.1	42	32
YOLOv3	Redmon et al. [4]	51.5	20	106
YOLOv3-tiny	Adarsh et al. [43]	33.1	220	24
YOLOv4	Bochkovskiy et al. [44]	43.5	65	137
YOLOv5	Jocher et al. [45]	48.1	264	24
YOLOv5s	Jocher et al. [45]	36.8	455	17
		Speed up	Test time per image (s)
R-CNN	Saeidi and Ahmadi [36]	1×	47
Faster R-CNN	Ren et al. [37]	146×	0.32