At this point, we’ve covered a lot of ground about how deep neural networks can help us understand image features and perform deterministic tasks on them, like object classification and detection. Now it’s time to turn our focus to a different, slightly more advanced area of computer vision and deep learning: generative models. These neural network models actually create new content that didn’t exist before--new people, new objects, a new reality, like magic! We train these models on a dataset from a specific domain, and then they create new images with objects from the same domain that look close to the real data. In this part of the book, we’ll cover both training and image generation, as well as look at neural transfer and the cutting edge of what’s happening in visual embeddings.