Chapter 1, Machine Learning Fundamentals, introduces the CRISP-DM model, which presents an industry standard framework/workflow for any data science, ML, or deep learning project. We will also touch upon various important concepts covering the fundamentals in the ML landscape such as exploratory data analysis, feature extraction and engineering, evaluation metrics, and so on.
Chapter 2, Deep Learning Essentials, provides a whirlwind tour of deep learning essentials, providing an overview of the basic building blocks of neural networks and also how deep neural networks are trained. Starting from the basics of how a single neural unit works, important concepts like activation functions, loss functions, optimizers and neural network hyperparameters are covered. Special focus is also emphasized on setting up on-premise and cloud-based deep learning environments.
Chapter 3, Understanding Deep Learning Architectures, focuses on understanding the various standard model architectures present today in deep learning. We have come a long way since traditional ANNs in the 1960s, and essential model architectures such as fully connected deep neural networks (DNNs), Convolutional Neural Networks (CNNs), recurrent neural networks (RNNs), Long-Short Term Memory (LSTM) networks, and the most recent Capsule Networks will be covered, to name a few.
Chapter 4, Transfer Learning Fundamentals, looks at the core concepts, terminology, and model architectures associated with the concept of transfer learning. Concepts and architectures pertaining to pretrained models will be discussed in detail. We will also compare and contrast transfer learning with deep learning and talk about types and strategies of transfer learning.
Chapter 5, Unleashing the Power of Transfer Learning, takes an actual example with a dataset from Kaggle, leverages deep learning models on it, and gives readers an understanding of the challenges faced when we have a small number of data points, and how transfer learning can unleash its true power and potential to give us superior models in these scenarios. We will tackle the very popular dogs and cats classification task here with the twist of a less data availability constraint.
Chapter 6, Image Recognition and Classification, is the first in a series of real-world applications/case studies of concepts discussed in detail in the previous two parts of the book. The chapter begins with an introduction to the task of image classification, and goes on to discuss and implement some of the popular, state-of-the-art deep learning models on diverse image classification problems.
Chapter 7, Text Document Categorization, discusses the application of transfer learning to a very popular natural language processing problem, text document categorization. The chapter begins with a high-level introduction to the multi-class text classification problem, traditional models, benchmark text classification datasets such as 20-newsgroups and performance. Later, it introduces deep learning document models for text classification and their advantages over traditional models. We learn about word feature representation using dense vectors and how to leverage the same for applying transfer learning in our text categorization problem such that our source and target domains might be different. Other unsupervised tasks like document summarization are also depicted.
Chapter 8, Audio Identification and Classification, solves the tough problem of identifying and classifying very short audio clips. Here we leverage transfer learning using some innovation of applying the power of pre-trained deep learning models from the computer vision domain into a totally different domain of audio identification.
Chapter 9, DeepDream, focuses on a gentle introduction to the domain of generative deep learning, which is one of the core ideas at the forefront of true artificial intelligence. We will be focusing on how convnets (CNNs) think or dream and visualize patterns in images by leveraging transfer learning. First released by Google in 2015, it became a viral sensation due to the interesting patterns deep networks started to generate from images, as if thinking and dreaming on their own!
Chapter 10, Style Transfer, leverages concepts from deep learning, transfer learning, and generative learning to showcase artistic image neural style transfer with hands-on examples on different content images and styles.
Chapter 11, Automated Image Caption Generator, covers one of the most complex problems in computer vision, as well as natural language generation—image captioning. While classifying images into fixed categories is challenging yet not impossible, this is a slightly more complex task involving generating human-like natural language textual captions for any photo or scene. Leveraging the power of transfer learning, natural language processing and generative models, you will learn how to build your own automated image captioning system from scratch.
Chapter 12, Image Colorization, presents a unique case-study where the task is to colorize black and white or grayscale images. This chapter introduces readers to the basics of various color scales and why image colorization is such a difficult task.
