Convolution layer

Parameters (kernel size, channel)

Feature output (high, wide)

Conv1(C1)

{(3 × 3, 32), (1 × 1, 16)}

h = H/2, w = W/2 (C1_2)

Max_pooling1

2 × 2

h = H/4, w = W/4

Conv2(C2)

{(3 × 3, 32), (1 × 1, 32)}

h = H/4, w = W/4 (C2_2)

Max_pooling2

2 × 2

h = H/8, w = W/8

Conv3(C3)

{(3 × 3, 64), (1 × 1, 32), (3 × 3, 64)}

h = H/8, w = W/8 (C3_3)

Max_pooling3

2 × 2

h = H/16, w = W/16

Conv4(C4)

{(3 × 3, 128), (1 × 1, 64), (3 × 3, 128)}

h = H/16, w = W/16 (C4_3)

Max_pooling4

2 × 2

h = H/32, w = W/32

Conv5(C5)

{(3 × 3, 256), (1 × 1, 128), (3 × 3, 256)}

h = H/32, w = W/32 (C5_3)

Max_pooling5

2 × 2

h = H/64, w = W/64

Conv6(C6)

{(3 × 3, 256), (1 × 1, 256), (3 × 3, 256)}

h = H/64, w = W/64 (C6_3)