A convolution within a neural network is a mathematical method that analyze images and process in a especific data design to process pixel.
It is used to extract group and filter in different shapes: lines, circles, squares, in any axe. Then we can use this shapes to clasify in a especific characteristic and crete a new model.
