Real-Time Human Ear Detection Based on the Joint of Yolo and RetinaFace

Table 1

DarkNet-53 architecture.

	Type	Filters	Size	Output
	Convolutional	32
	Convolutional	64
	Convolutional	32
	Convolutional	64
	Residual
	Convolutional	128
	Convolutional	64
	Convolutional	128
	Residual
	Convolutional	256
	Convolutional	128
	Convolutional	256
	Residual
	Convolutional	512
	Convolutional	256
	Convolutional	512
	Residual
	Convolutional	1024
	Convolutional	512
	Convolutional	1024
	Residual
	Avg pool		Global
	FC		1000
	Softmax