Shulou(Shulou.com)06/03 Report--

Http://product.dangdang.com/25207334.html

Brief introduction of content

The general guiding ideology of this book is to cultivate the ability to use TensorFlow for practical programming to solve problems related to image processing on the basis of mastering the basic knowledge and characteristics of deep learning. The book strives to be profound and simple, introducing the basic usage of TensorFlow, high-level model design and corresponding programming through easy-to-understand language and detailed program analysis.

This book consists of 22 chapters, including the installation and use of the Python class library, the basic data structure and use of TensorFlow, the creation and reading of TensorFlow data sets, artificial neural networks, feedback neural networks, the theoretical basis of full convolution neural networks, the creation of deep learning models, model characteristics, algorithms, ResNet, Slim, GAN and so on. This book emphasizes the combination of theory with practice, focuses on the application of TensorFlow programming to solve image recognition, provides a large number of data sets, and implements a deep learning model in the form of code for readers' reference.

This book can be used not only as a training and self-study book for programmers to study artificial neural networks, deep learning, TensorFlow programming and image processing, but also as teaching materials for related majors in colleges and universities and training institutions.

Foreword

We are in an era of change!

Given an object, it seems very simple to ask a 3-year-old child to describe what the object is. However, it was impossible not long ago to put the same thing in front of the computer and let it describe what it saw.

Let computers learn to "see" things, this is a special subject-- computer vision is doing. With the development of artificial neural network and deep learning, a great breakthrough has been made in the research of computer vision in recent years. The convolution neural network model constructed by simulating biological vision has achieved very good results in image recognition and classification.

Now, with the development of deep learning, the use of artificial intelligence to handle routine labor, understand phonetic semantics, help medical diagnosis and support basic scientific research seems to be a dream.

The reason for writing this book

As the latest and most widely used open source framework for deep learning, TensorFlow has naturally attracted wide attention. It attracts a large number of programmers and developers to develop and learn related content. Programmers who master the basic skills of TensorFlow programming have become hot talents eagerly sought by various organizations and units. Their main work is to use the obtained data set to design different artificial neural models, use the strong learning ability of artificial neural network to extract and mine the potential information contained in the data set, write corresponding TensorFlow programs to deal with the data, and further develop its value, so as to provide corresponding support for the acquisition of business opportunities, innovation of management models, and decision-making. As more and more organizations, units and industries attach importance to the application of deep learning, high-level TensorFlow programmers are bound to become tight talents in the job market.

At present, TensorFlow has been published by Google for less than two years, but it is widely used in industrial, commercial and scientific research, making it the most popular deep learning framework nowadays. Due to the lag of domestic translation and knowledge dissemination and other reasons, there is a lack of domestic introduction in this area, and the lack of relevant content about the use and design of the latest TensorFlow framework, resulting in the delay of knowledge dissemination. Learning is to master new knowledge and acquire new abilities, not to learn what has been abandoned.

Secondly, different from other application frameworks, TensorFlow is not a simple programming framework, and deep learning is not a simple term, but requires relevant researchers to learn the theory behind its code and master some mathematical knowledge and theoretical basis. The author has a long-term basis of front-line science theory teaching, and the theoretical knowledge can be introduced and described in a very easy-to-understand language, which is incomparable to some relevant books on the market.

This book is published to meet the requirements of TensorFlow programmers and developers to learn the latest TensorFlow program code. This book introduces and explains the structure and programming code of deep learning step by step, based on solving the actual image processing, introduces the TensorFlow programming pattern from the theory, provides a full and accurate analysis of its principle and implementation from many angles and aspects, and combines the application design written with practical cases, so that readers can master TensorFlow programming methods and skills at the developer's level. It lays a solid foundation for the development of more powerful image processing applications.

The advantages of this book

L this book focuses on the use of convolution neural network and its related change model in the direction, image feature extraction, image recognition and specific applications on the TensorFlow framework, which is rarely involved in the market.

This book is not a boring theoretical explanation, but a summary of a great deal of the latest literature. In this respect, this book is fundamentally different from other programming books. The examples in this book come from the winning models of image resolution and feature competition in the real world. By introducing these examples, readers can further understand and master their inherent algorithms and nature.

The author of this book has long-term graduate and undergraduate teaching experience, explains all the contents in an easy-to-understand language, introduces all the knowledge points of the theoretical system of feedback neural network and convolution neural network, and uses the latest framework of TensorFlow recommended by the government to design programs to help readers better use the latest model framework to understand and master the subtleties of TensorFlow programming.

The author believes that people who master and use in-depth learning should pay attention to the cultivation of practical application development ability and problem-solving ability on the basis of mastering basic knowledge and theory. Therefore, this book analyzes a large number of practical cases encountered by the author in practical work, abstracts the core model and gives specific solutions, and all program examples provide corresponding codes for readers to learn.

The content of this book

The book is divided into 22 chapters, all of which are written in Python, which is the language recommended by the TensorFlow framework.

Chapter 1 introduces the basic contents of deep learning, introduces the application and development direction of deep learning in computer vision, and introduces the application prospect of using deep learning to solve computer vision problems. The purpose of this paper is to show that the use of deep learning and artificial intelligence to achieve computer vision is the future development direction, but also the inevitable trend.

Chapter 2 introduces the installation of Python and the most commonly used class libraries. Python is a very easy-to-use language that can easily express formulas and visions in the form of code without having to learn too much programming knowledge. Python special class library threading is not common, just to lay the foundation for later data reading and TensorFlow special format generation.

Chapter 3 comprehensively introduces the basic classification, algorithm and theoretical basis of machine learning, and introduces the specific implementation and application of different algorithms, such as regression algorithm and decision tree algorithm. These are the basic theoretical parts of deep learning, which thoroughly and accurately show the structure and application of deep learning to readers, and lay a solid foundation for further mastering the application of deep learning in computer vision.

Chapter 4 mainly introduces the use of Python language. Through the introduction and implementation of different Python class libraries, it helps readers to strengthen the programming ability of Python and learn the corresponding class libraries. These are the contents that are used repeatedly in the following text. At the same time, borrow the acquired knowledge to learn the visual display skills of data. Although this skill is a basic skill in data analysis, it plays a very important role.

Chapter 5 ~ 6 is an introduction to the use of OpenCV class library. This book focuses on image processing, and the reading, editing and processing of image data is the top priority of this book. OpenCV is a class library specially used for image processing in Python. Through basic explanation and advanced introduction, readers can master the use of this important class library. Learn to write code for image clipping, transformation and translation. Chapter 5 introduces the basic content of convolution kernel in the form of an example, and realizes the function of convolution kernel in Python language. Convolution kernel is not only a very important basic part of this book, but also a very important part of image processing. It is very helpful to write corresponding programs to achieve convolution check image processing, master and understand convolution neural network.

Chapter 7-8 is the basic foundation of TensorFlow, which shows readers the basic application of TensorFlow through an entertainment website, demonstrates the fitting process of neural network classification by means of graphics and images, and understands the content behind it while entertaining.

Chapter 9 is a key point of this book and the basic content of neural network. The feedback algorithm in this chapter is a milestone algorithm to solve the excessive computational complexity of neural network. Through a detailed and careful explanation, the author introduces the algorithm in an easy-to-understand language, and realizes the most important algorithm content in this neural network for readers by writing code independently. The content of this chapter does not seem much, but it is very important.

Chapter 10 gives a detailed introduction to the data input and output of TensorFlow. From reading CSV files to teaching readers to make a special TensorFlow data format TFRecord, this is rarely mentioned in the books on the market at present. Using the TensorFlow framework for programming, data preparation and standardization is a top priority, so this chapter is also a more important chapter.

Chapter 11 ~ 12 is a basic course of using convolution neural network to learn on the TensorFlow framework. After the preparation and introduction of the previous chapter, using the basic theory-convolution neural network for handwriting identification is the most basic skill of deep learning, and it is also a very important learning basis. And in the process of programming, the author shows the reader that parameter adjustment plays an important role in the model test results. This is the content that is not involved in the relevant books on the market, and it is very important.

Chapter 13 ~ 14 is the introduction and application of convolution neural network algorithm. In these two chapters, the author introduces in detail the application of convolution neural network, especially in image recognition, from handwritten numerical recognition to display object recognition. With the help of the data set of the image recognition competition, the convolution neural network model which won the prize in the competition is used to enable the reader to master the variation of the convolution neural network. The theoretical basis of convolution neural network is the forward and reverse process of convolution. Generally, the forward process is easy to understand and learn, but for reverse operation, it is basically not involved, and in some cases, it is only the copy and excerpt of the formula. In this book, the operation and calculation method of the reverse process of convolution neural network are introduced in detail in chapter 14. Through the description of a large number of examples, the reverse operation of convolution neural network is described in detail for the first time. This is what is missing in the relevant books.

Chapter 15 demonstrates the process of using convolution neural network for image recognition through a complete example. The example comes from the ImageNet image recognition competition, and the model used is also the model with the highest accuracy in the competition. Through the detailed analysis of each step of the project, the reader is taught hand in hand how to use convolution neural network for image recognition.

Chapter 16 introduces the structure of VGGNet, focusing on the network tuning parameters of VGGNet and the ability to execute Finetuning after that. This chapter reuses the VGG16 implementation of the examples in Chapter 15 to provide readers with a way to solve the problem from different perspectives and different model methods.

Chapter 17 aims at the answers to some interview questions given by the current deep learning workers, which can not only help recruiters analyze who is a high-level interviewer, but also help them improve their technical concepts and knowledge and find their own positions. pave the way for promotion and salary increase in the future.

Chapter 18 introduces the deep learning network ResNet model, which uses a large number of residual modules as the basic component of the network. The main function is to make the network increase with the change of depth without weight attenuation and gradient attenuation or disappearance. In addition to the ResNet model, this chapter also introduces the new convolution neural model, including SqueezeNet and Xception.

Chapters 19-20 enter the advanced stage of TensorFlow learning, focusing on an API--Slim, a lightweight development library for defining, training, and evaluating more complex models. These two chapters not only introduce how to use it, but also make a multi-layer perceptron MLP, a convolution neural network CNN, and finally use Slim pre-training model for Finetuning.

Chapter 21 introduces full convolution neural network image segmentation, first explains the theoretical basis and implementation method of image segmentation, then gives the step-by-step flow and programming basis of full convolution neural network for image segmentation, and finally gives the actual practice of image segmentation using VGG16 full convolution network.

Chapter 22 explains the GAN-- antagonistic generation network. although the theory of this chapter seems boring, the author uses a "generator" and a "discriminator" to carry out "confrontation" in a network to reduce the difficulty of reading. Finally, through the case of using GAN to generate handwritten numbers, readers can really learn the application of GAN.

In addition, the book introduces the deep learning model which is the most popular and achieves the best results in image recognition, which is the focus and research focus of deep learning at present.

The characteristics of this book

This book is not a pure introduction of theoretical knowledge, nor an advanced technical discussion, but entirely from practical application, extending core knowledge with the simplest and typical examples, and finally pointing out the way to further in-depth study of "advanced and sophisticated".

This book does not introduce a certain knowledge block in depth, but comprehensively introduces the basic structure and upper programming design of image processing involved in TensorFlow, systematically and comprehensively explains the whole picture of deep learning, so that readers can grasp the direction in the process of learning.

L this book is easy to understand in writing, there is no esoteric mathematical knowledge, but uses a more vivid form, using a large number of image examples to describe the theoretical knowledge of the application, so that readers can master the relevant content under easy and pleasant reading.

The purpose of this book is to guide readers to make more technical innovations, and each chapter will help readers better understand the content of this chapter in the form of sample descriptions.

The code of this book follows the principle of refactoring, avoids code pollution, and sincerely hopes that readers can write excellent, concise and maintainable code.

This book is suitable for people.

The download address of the sample source code for this book (pay attention to the case of numbers and letters) is as follows:

Https://pan.baidu.com/s/1jHFg2uq

If you have problems with downloading or have any questions about this book, please contact booksaga@163.com with the subject of email "TensorFlow".

This book is suitable for people.

This book is not only suitable for programmers who study artificial neural networks, deep learning and TensorFlow programming, but also can be used as teaching materials for related majors in colleges and universities. It is suggested that in the process of learning the content of this book, theory should be combined with practice, some codes should be compiled independently, and an open experimental method should be adopted, that is, readers prepare experimental data and experimental environment by themselves to solve practical problems, and finally achieve the goal of combining theory with practice.

The author of this book

The author of this book is currently a computer professional teacher, responsible for data mining, Java programming, data structure and other undergraduate and graduate courses, research interests for data warehouse and data mining, artificial intelligence, machine learning, research and participation in a number of scientific research projects. I would like to thank my family and friends for their strong support in the process of writing.

Wang Xiaohua

November 2017

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.