Hands-On Automated Machine Learning

What are filters?

A CNN uses filters to identify features in an image, such as edges, lines, or arches. They search for certain important patterns or features in an image. Filters are designed to search for certain characteristics in an image and detect whether or not the image contains those characteristics. A filter is applied at different positions in an image, until it covers the whole image. Filters form a critical element in a convolution layer, an important step in a CNN.

There are mainly four different layers in a CNN:

• Convolution layer
• ReLU layer
• Pooling layer
• Fully connected layer

Let's discuss the convolution layer, which is the first stage in a CNN.