Research on Object Detection Algorithm Based on Multilayer Information Fusion

Table 1

The structural parameters of DenseNet-98.

Layers	Output size	DenseNet-98
Convolution	224 × 224 (d = 16)	7 × 7 conv, stride = 2
Pooling	112 × 112 (d = 16)	3 × 3 max pool, stride = 2
Dense block (1)	112 × 112 (d = 160)	[1 × 1 conv + 3 × 3 conv] × 6
Transition	112 × 112 (d = 80)	1 × 1 conv
Layer (C1)	56 × 56 (d = 80)	2 × 2 average pool, stride = 2
Dense block (2)	56 × 56 (d = 368)	[1 × 1 conv + 3 × 3 conv] × 12
Transition	56 × 56 (d = 184)	1 × 1 conv
Layer (C2)	28 × 28 (d = 184)	2 × 2 average pool, stride = 2
Dense block (3)	28 × 28 (d = 472)	[1 × 1 conv + 3 × 3 dilated-conv] × 12
Transition	28 × 28 (d = 256)	1 × 1 conv
Layer (C3)
Dense block (4)	28 × 28 (d = 640)	[1 × 1 conv + 3 × 3 dilated-conv] × 16
Transition	28 × 28 (d = 256)	1 × 1 conv
Layer (C4)