ResNet15: Weather Recognition on Traffic Road with Deep Convolutional Neural Network

Table 2

Architecture of ResNet50 and ResNet15.

Layer name	ResNet50			ResNet15
	Shape		Output	Shape		Output
Conv1	[7 × 7]:64/2		[112 × 112]:64	[7 × 7]:32/2		[112 × 112]:32
MaxPooling	[3 × 3]/2		[56 × 56]:64	[3 × 3]/2		[56 × 56]:32
Conv2_x	[1 × 1]:64	CB×1/1		[1 × 1]:32
	[3 × 3]:64		[56 × 56]:256	[3 × 3]:32	CB×1/2	[28 × 28]:64
	[1 × 1]:256	IB×2/1		[1 × 1]:64
Conv3_x	[1 × 1]:128	CB×1/2		[1 × 1]:64
	[3 × 3]:128		[28 × 28]:512	[3 × 3]:64	CB×1/2	[14 × 14]:128
	[1 × 1]:512	IB×3/1		[1 × 1]:128
Conv4_x	[1 × 1]:256	CB×1/2		[1 × 1]:128
	[3 × 3]:256		[14 × 14]:1024	[3 × 3]:128	CB×1/2	[7 × 7]:256
	[1 × 1]:1024	IB×5/1		[1 × 1]:256
Conv5_x	[1 × 1]:512	CB×1/2		[1 × 1]:256
	[3 × 3]:512		[7 × 7]:2048	[3 × 3]:256	CB×1/2	[4 × 4]:512
	[1 × 1]:2048	IB×2/1		[1 × 1]:512
Others	AveragePooling, [7 × 7]/7		[1 × 1]:512	FC-512	512
	Softmax-4		4	Softmax-4	4
Model size	45.7M			5.4M