Building a CNN
CNNs are the ideal architecture when dealing with an image data problem. However, they are often underused as they are typically used for image classification tasks, even though their abilities extend to other domains in the field of image processing. This chapter will not only explain the reasons why CNNs are so good at understanding images but will also identify the different tasks that can be tackled, as well as provide some examples of real-life applications.
Moreover, this chapter will explore the different building blocks of CNNs and their application using PyTorch to ultimately build a model that solves a data problem using one of PyTorch's datasets for image classification.
Why Are CNNs Used for Image Processing?
An image is a matrix of pixels, so why not just flatten the matrix into a vector and process it using a traditional neural network architecture? The answer is that, even with the simplest image, there are some pixel dependencies that alter...
