Convolutional neural networks (CNN) are similar to the multi-layer perceptron network. They are ideal for computer vision applications, although it can be used for non-image applications. In contrast to a multi-layer perceptron, CNNs extract features through a simple convolutional operation. The hidden layers consist of kernels or filters that move across the input image and perform
the convolutional operation. This mathematical operation essentially computes the element-wise multiplication between the kernel matrix and the input matrix, and the resultant values are summed. 

 

Let's dive into the code: