Skip to main content

Table 1 Adaptability improvement of the PrimaryCaps layer

From: Dynasty recognition algorithm of an adaptive enhancement capsule network for ancient mural images

Parameter CapsNet AEC
Input feature map 20 × 20 × 1 28 × 28 × 256
Number of convolution kernels 256 256
Kernel size 9 × 9 × 1 3 × 3 × 256
Step of convolution movement 2 2
Size of the output feature map 6 × 6 × 256 14 × 14 × 256
Number of the output features 9216 50,176
Number of the sealed capsules 1152 6272
Size of the sealed capsules 8 × 1 8 × 1