Embedded YOLO: A Real-Time Object Detector for Small Intelligent Trajectory Cars

Table 4

Feature fusion network comparison experiment (PANet vs. PANet-Tiny).

Neck network	Backbone network	Parameters (M)	Accuracy (%)
PANnet [16]	ShuffleNet v2 [22]	14.33	21.7
Our PANet-Tiny		3.71	21.7
PANnet [16]	Our ASU-SPP	12.62	23.3
Our PANet-Tiny		3.25	23.3
PANnet [16]	MobileNetV3 [24]	11.8	22.9
Our PANet-Tiny		4.84	22.9
PANnet [16]	ThunderNet [25]	10.96	23.2
Our PANet-Tiny		4.52	23.2