Deep Learning for Person Reidentification Using Support Vector Machines

Table 1

The layer parameters of our network. The output dimension in the table is given by height × width × width. All convolution and fully connected layers use Relu activation function.

Name	Output dim	Filter size	Stride
C1	55 × 55 × 96	11 × 11	4
S1	27 × 27 × 96	3 × 3	2
C2	27 × 27 × 256	5 × 5	1
S2	13 × 13 × 256	3 × 3	2
C3	13 × 13 × 384	3 × 3	1
C4	13 × 13 × 384	3 × 3	1
C5	13 × 13 × 256	3 × 3	1
S5	6 × 6 × 256	3 × 3	2
F6	—	4096	—
F7	—	4096	—
L8	2	—	—