CNN KERNEL : Performance analysis based on kernel size of Convolutional Layers in a network

In this book I perform an experimental review on twelve similar types of Convolutional Neural Network architecture but the different sizes of kernels for the filters.  For this experiment I select twelve different sizes of the kernel for twelve Convolutional Neural Network models the size of kernels are – (12 12) (11 11) (10 10) (9 9) (8 8) (7 7) (6 6) (5 5) (4 4) (3 3) (2 2) and (1 1).  For this experiment I use the “Flowers Recognition” dataset. I use 77 batches (batch size = 45) per epoch and 10 epochs per experimental fold. After analyzing the results I found that according to the performance kernel_size (2 2) and (3 3) are the best selection for the two-dimensional convolutional layer in the convolutional neural networks. The goal of this experiment is to help the developer to understand and select the perfect size of the kernel for filter during two-dimensional image processing by using the two-dimensional Convolutional (Conv2D) layer [11] of CNNs.