Summary

In this chapter, we have seen how to use and build real-life applications using CNNs, which are a type of feedforward artificial neural network in which the connectivity pattern between neurons is inspired by the organization of the animal visual cortex. Our image classifier application using CNN can classify real-life images with an acceptable level of accuracy, although we did not achieve higher accuracy. However, readers are encouraged to tune hyperparameters in the code and also try the same approach with another dataset.

Nevertheless, and importantly since the internal data representation of a convolutional neural network does not take into account important spatial hierarchies between simple and complex objects, CNN has some serious drawbacks and limitation for certain instances. Therefore, I would suggest you take a look at the recent activities around capsule networks on GitHub at https://github.com/topics/capsule-network. Hopefully, you can get something useful out from there...