序三

Among many books about OpenCV that have been published already,this one has some interesting properties that make it special.It has been written by the members of Intel China team who optimized OpenCV deep learning module(OpenCV DNN)for GPU.So,the book contains some in-depth first-hand information that would be difficult to find somewhere else.That includes some rarely discussed topics,such as OpenCL,Vulkan and Halide backends of OpenCV DNN.Another expert-level and very useful topic is discussion with practical examples of how to analyse application performance and then tune it using Intel VTune.At the same time,the book is not just for experts.It also provides a comprehensive guide to OpenCV DNN usage,from the simple models,like image classification,to more complex models,such as object detection.Overall,the book is highly recommended for all the software engineers who want to build highly-efficient deep learning applications using OpenCV.

在已经出版的众多关于OpenCV的书籍中，本书具有一些有趣的特性，因之与众不同。本书是由Intel中国团队的成员编写的，他们优化了OpenCV模块在GPU上的性能。因此，本书包含了一些深度的第一手信息，这些信息在其他地方很难找到。其中包括一些很少讨论的有用话题，如OpenCL、Vulkan，以及OpenCV DNN的Halide后端。另一个专家级且非常有用的主题是讨论如何使用Intel VTune分析应用程序性能，然后根据它提供的信息对程序进行优化。本书的读者对象包含但不限于专家。它为OpenCV DNN模块的使用提供了全面的指导，从简单的模型（如图像分类）到更复杂的模型（如对象检测）。总体来说，本书值得强烈推荐给所有希望使用OpenCV构建高效深度学习应用程序的软件工程师。

Vadim Pisarevsky

OpenCV team lead

Senior Software Engineer,AIRS(Artificial Intelligence for Robotics and Society)Institute,Shenzhen

In Russia