CNNs
CNNs share many common components with the ANNs you have built so far. The key difference is the inclusion of one or more convolutional layers within the network. Convolutional layers apply convolutions of input data with filters, also known as kernels. Think of a convolution as an image transformer. You have an input image, which goes through the CNN and gives you an output label. Each layer has a unique function or special ability to detect patterns such as curves or edges in an image. CNNs combine the power of deep neural networks and kernel convolutions to transform images and make these image edges or curves easy for the model to see. There are three key components in a CNN:
- Input image: The raw image data
- Filter/kernel: The image transformation mechanism
- Output label: The image classification
The following figure is an example of a CNN in which the image is input into the network on the left-hand side and the output is generated on the right-hand side...
