Research Article

Pavement Crack Detection and Segmentation Method Based on Improved Deep Learning Fusion Model

Table 1

SSD convolution feature layers and feature map outputs.

VGG16 layer (type)Resnet layer (type)Output shape

conv4_3conv2d_24(64, 64, 512)
conv2d_43(32, 32, 1024)
conv6_2conv2d_53(16, 16, 512)
conv7_2conv2d_63(8, 8, 256)
conv8_2conv2d_73(4, 4, 256)
conv9_2conv2d_83(2, 2, 256)
conv10_2add_28(1, 1, 256)