In this chapter, we covered the basics of similarity learning. We studied algorithms such as metric learning, Siamese networks, and FaceNet. We also covered loss functions such as contrastive loss and triplet loss. Two different domains, ranking and recommendation, were also covered. Finally, the step-by-step walkthrough of face identification was covered by understanding several steps including detection, fiducial points detections, and similarity scoring.
In the next chapter, we will understand Recurrent Neural Networks and their use in Natural Language Processing problems. Later, we will use language models combined with image models for the captioning of images. We will visit several algorithms for this problem and see an implementation of two different types of data.