|
|
|
|
移动端

序言

《深度学习与计算机视觉:算法原理、框架应用与代码实现》本书全面介绍了深度学习及计算机视觉中最基础的知识,并结合最常见的应用场景和大量实例,带领读者进入丰富多彩的计算机视觉领域。本节为序言。

作者:叶韵来源:机械工业出版社|2017-11-16 16:35

序言

深度学习是机器学习的一个重要分支,它以简化的方式模拟人脑复杂的神经系统,从而达到对数据的高级抽象。近些年,深度学习在语音识别、计算机视觉、自然语言处理、生成网络和无监督学习等领域都有着广泛的应用,从很多方面改变着人们的日常生活。

互联网巨头谷歌、Facebook、亚马逊、微软、百度、阿里巴巴和腾讯等公司都建立了相应的深度学习部门和平台。随着近几年深度学习的快速发展,相继出现了大量的开源软件平台,如Caffe、MXNet、TensorFlow和Torch等。这些平台多数都有相应的Python和C++接口,功能非常强大。但是对于初学者来说,还是有一定的门槛。

本书架起了一座初学者和开源深度学习软件之间的桥梁,致力于帮助初学者进入机器学习特别是深度学习在计算机视觉中的应用等领域。本书涵盖了基础的数学、机器学习和图像识别等内容,同时对两个主流的开源深度学习库Caffe和MXNet都有大量的实战例子描述分类和回归等问题。

本书作者在深度学习领域有着深入的研究,善于把复杂的问题用浅显易懂的语言描述出来,使得本书内容引人入胜。本书结构合理,内容涵盖了计算机视觉领域的一些主要问题。对于一个学习计算机视觉的新手来说,本书的数学推导浅显易懂,从一些简单例子开始,然后推广到抽象的矩阵描述方式,大大减轻了学习负担。读者可以通过前7章的学习,对神经网络、基础的数学和编程技巧有一个全面的了解。在此基础上可以根据具体的问题参考本书中具体的章节,例如图像识别、回归和目标检测等。在这些章节中,读者可以按照书中的步骤搭建自己的应用。

田疆

西门子高级研究员

附英文原文

Deep learning is a branch of machine learning based on a set of algorithms that attempt to model high level abstractions in data, and mimics human brain’s neural networks in a simplistic manner. It is powerful and driving advances in speech recognition, computer vision, natural language processing, generative networks, and unsupervised machine learning in recent years, which are changing our daily life from different aspects.

Internet giants Google, Facebook, Amazon, Microsoft, Baidu, Alibaba, and Tencent launch their deep learning teams and platforms. With the rapidly growing related research and development community, in recent years, there appear plenty of open source frameworks for deep learning, such as: Caffe, MXNet, TensorFlow, Torch, etc. These frameworks can usually interface with Python, Lua, and C++. They are powerful and usually well maintained, however, they are not that straightforward and "easy cooking recipes" for beginners.

This book bridges the gap with the spirit of open source. It is a resource intended to help students and practitioners enter the field of machine learning in general and deep learning (computer vision) in particular. This book not only covers basic theory on math, machine learning, image recognition, but also helps deep learning practitioner on two major libraries Caffe and MXNet by plenty of examples on classification, regression, metric learning, etc. The tutorials in this book also provide step-by-step instructions for creating models for specific types of applications.

The writing of this book is good and also interesting, which shows the author's delving very deep into this field. The structure of this book is quite reasonable. Its topics cover the major tasks in computer vision. The mathematics derivation in this book is solid and understandable for a computer vision novice, it starts from simple examples, and then to more abstract matrix counterparts, which greatly reduces the "activation threshold" for a learning curve.

Readers of this book may start the first seven chapters to get a comprehensive understanding of neural network, basic math and programming techniques. Thereafter, he may pick up the related applications such as image recognition, regression, fine-tuning, image detection, metric learning, and neural art in the following chapters, wherein, he can get step by step instructions to build the demo, which is "train by example" learning.

Jiang Tian, PHD

Lead Research Scientist, Siemens


喜欢的朋友可以添加我们的微信账号:

51CTO读书频道二维码


51CTO读书频道活动讨论群:365934973

【责任编辑:book TEL:(010)68476606】

回书目   上一节   下一节
点赞 0
分享:
大家都在看
猜你喜欢

读 书 +更多

SOA概念、技术与设计

在本书中,Thomas ERL呈现了第一部端对端的教程,提供了从基层开始的面向服务的建模与设计的逐步指导。通过逐步的、清晰生动的、良好的SOA...

订阅51CTO邮刊

点击这里查看样刊

订阅51CTO邮刊