Pretraining Convolutional Neural Networks for Image-Based Vehicle Classification

Table 1

The structure of C4M3F2.

Layer Type/Activation	Size/Stride	Filters
Convolutional/ReLU	3×3/1	32
Max Pooling	2×2/2
Convolutional/ReLU	3×3/1	64
Convolutional/ReLU	3×3/1	64
Max Pooling	2×2/2
Convolutional/ReLU	3×3/1	64
Max Pooling	2×2/2
Fully Connected/ReLU	1024
Fully Connected	1024
Softmax	4