Introduction
A filter (or kernel) is a small grid of numbers, for example 3 × 3. A convolutional layer slides this filter over the image. At each position, it multiplies the filter's numbers by the pixels underneath, adds them up, and writes the result into an output grid. That output grid is called a feature map: it shows where in the image the filter's pattern appears, and how strongly.
Think of a stencil for a particular shape. You slide it across a picture; wherever the picture matches the stencil's shape well, you mark a bright spot. When you're done, the marks form a map of "where this shape is". Each filter is a different stencil, and the network learns its own stencils.