Research on Road Adhesion Condition Identification Based on an Improved ALexNet Model

Table 1

ALexNet structure parameters.

Name	Parameter	Stride	Number of convolution kernels
Conv 1	11 × 11 × 3	4	96
Pool 1	1 × 3 × 3 × 1	2	—
Conv 2	5 × 5 × 48	1	256
Pool 2	1 × 3 × 3 × 1	2	—
Conv3	3 × 3 × 256	1	384
Conv4	3 × 64 × 384	1	384
Conv5	3 × 3 × 384	1	256
Pool5	1 × 3 × 3 × 1	2	—
Fc6	4096	—	—
ReLU6	ReLU	—	—
Dropout	0.5	—	—
Fc7	4096	—	—
ReLU7	ReLU	—	—
Dropout	0.5	—	—
Fc8	1000	—	—
Prob	SoftMax	—	—
Output	1000	—	—