In this chapter, we've looked at using neural networks to classify images. It's very different from working with normal data. Not only do we need a lot more training data to get the right result, we also need a different architecture to work with images that is better suited for the job.
We've seen how convolution layers and pooling layers can be used to essentially create an advanced photo filter that extracts important details from the data and summarize these details to reduce the dimensionality of the input to a manageable size.
Once we have used the advanced properties of the convolution filters and pooling filters, it's back to business as usual with dense layers to produce a classification network.
It can be quite hard to come up with a good structure for an image classification model, so it's always a good idea to check out one of the...
