Скачать или смотреть 51- Concept of Filter/ Kernel in Convolutional Neural Network(CNN)

The command filters in Conv2D layer of keras represents the number of filters. Each filter slides over the input image, generating a “feature map” as output.
In image processing kernel is a convolution matrix or masks which can be used for blurring, sharpening, embossing, edge detection, and more by doing a convolution between a kernel and an image. As far as choosing the appropriate value for no. of filters, it is always recommended to use powers of 2 as the values. This filter is also sometimes called a Windows, or kernels.

Kernel_size is two-dimensional width and height of the filter.
This parameter determines the dimensions of the kernel. Common dimensions include 1×1, 3×3, 5×5, and 7×7 which can be passed as tuples.
The second required parameter you need to provide to the Keras Conv2D class is the kernel_size, a 2-tuple specifying the width and height of the 2D convolution window. The kernel size must be an odd integer as well.
It’s rare to see kernel sizes larger than 7×7.

So, when do you use each?
If your input images are greater than 128×128 you may choose to use a kernel size greater than 3 to help learn larger spatial filters and to help reduce volume size. Other networks, such as VGGNet, exclusively use 3x 3 filters throughout the entire network.