As mentioned, we will not focus on the math going on behind the scenes of face detection. Instead, we will use an external service to perform the recognition for us. The external service is MachineBox. What it does is quite clever. Instead of having to write your own deep learning algorithms, MachineBox packages up the commonly-used deep learning functionalities into containers, and you simply just use them straight out of the box. What do I mean by commonly-used deep learning functionalities? Nowadays people are relying more and more on deep learning for tasks such as facial recognition.

Just like Viola-Jones in the early 2000s, there are only a few commonly used models—we used the Haar-like cascades generated by Rainer Lienhart in 2002. The same is becoming true of deep learning models, and I shall talk more about the implications of that in the next chapter. By models, I mean the actual weights of the deep learning networks (for a more in-depth coverage, see Chapter 7, Convolutional Neural Networks – MNIST Handwriting Recognition, on deep neural networks). These commonly-used models are packaged up by MachineBox and you're able to just use it out of the box.

One thing that must be kept in mind is that MachineBox is a paid service. They do offer a free tier, which is sufficient for the needs of this chapter. I am in no way affiliated with MachineBox. I just think they're a cool company and deserve some recognition for the work they do. Plus, they do not do sketchy things such as secretly charging your credit card, so that's a plus from me.