Machine Learning for Computer Vision

Size: px

Start display at page:

Download "Machine Learning for Computer Vision"

祚辛昝
7 years ago
Views:

1 基于 MATLAB 的计算机视觉插上人工智能的翅膀 MathWorks China Application Engineer Tom Shan 2014 The MathWorks, Inc. 1

2 题纲基于 MATLAB 的人工智能设计流程介绍演示人工智能流程和常见挑战针对实时视频进行目标检测训练目标检测器并行计算加速集群 GPU/ 多 GPU 加速总结 3

3 题纲基于 MATLAB 的人工智能设计流程介绍演示人工智能流程和常见挑战针对实时视频进行目标检测训练目标检测器并行计算加速集群 GPU/ 多 GPU 加速总结 4

4 使用机器学习都能帮我们解决哪些问题? Clustering Object Recognition or Classification Object Detection 5

5 机器学习 Machine Learning 机器学习使用数据, 并以此产生出一套用于特定任务的程序任务 : Image Category Recognition Hand Written Program boats mugs hats Computer Vision Machine Learning boats mugs hats If brightness > 0.5 then hat If edge_density < 4 and major_axis > 5 then boat model = < Machine Learning >(data, label) Algorithm 6

Machine Learning boats mugs hats If brightness > 0.

6 题纲基于 MATLAB 的人工智能设计流程介绍演示人工智能流程和常见挑战针对实时视频进行目标检测训练目标检测器并行计算加速集群 GPU/ 多 GPU 加速总结 7

7 使用图像进行机器学习的流程 Training Data Feature Extraction Learning or Modelling Training Classifier / Model hat Input Image Feature Extraction Classification Classification 8

8 挑战 : 使用图像进行机器学习的流程 Training Data Feature Extraction Learning or Modelling Challenge 1 Challenge 2 Challenge 3 Challenge 2 Classifier / Model hat Input Image Feature Extraction Classification 9

9 基于图像机器学习的共同挑战 Challenge 1: Handling large sets of images Challenge 2: How to extract discriminative information from images Challenge 3: How to model tasks or data using machine learning 10

10 题纲基于 MATLAB 的人工智能设计流程介绍演示人工智能流程和常见挑战针对实时视频进行目标检测训练目标检测器并行计算加速集群 GPU/ 多 GPU 加速总结 11

11 目标 : 对实时视频中的目标进行识别 / 分类 vs vs hat mug boat Known as object classification or recognition 12

12 什么是特征提取? Bag of Words SURF HOG Image Pixels 特征提取图像特征通常是对于缩放旋转亮度的改变, 具有一致性比直接存储像素数据, 更加紧凑需根据实际问题, 进行特征的选取 Sparse Dense 13

13 Bag of Words Image Processing Toolbox Class / Label Perform image processing, analysis, and algorithm development Image Processing Toolbox provides a comprehensive set of reference-standard algorithms, functions, and apps for image processing, analysis, visualization, and algorithm development. You can perform image analysis, image segmentation, image enhancement, noise reduction, geometric transformations, and image registration. Many toolbox functions support multicore processors, GPUs, and C-code generation. Image Processing Toolbox supports a diverse set of image types, including high dynamic range, gigapixel resolution, embedded ICC profile, and tomographic. Visualization functions and apps let you explore images and videos, examine a region of pixels, adjust color and contrast, create contours or histograms, and manipulate regions of interest (ROIs). The toolbox supports workflows for processing, displaying, and navigating large images. Training Data Vocabulary / Bag of Words 14

You can perform image analysis, image segmentation, image enhancement, noise reduction, geometric transformations, and image registration.

14 Bag of Visual Words ( features) mugs Class / Label Training Data Vocabulary / Bag of Words 15

15 使用 Bag of Words 进行图像识别 / 分类 Training Data Bag = Visual Vocabulary Bag Classifier hat Input Image 16

16 Many Options for Features and Machine Learning Feature Extraction BRISK,FREAK, SURF Histogram of Oriented Gradients (HoG) Using box filters(integral images ) Bag of visual words Color-based features Frequency-domain features Machine Learning SVM Decision trees AdaBoost Bagged trees k-nn Discriminant analysis Bayes classifiers Bottom Line: Many permutations and combinations to fit the needs of your problem 19

Frequency-domain features Machine Learning SVM Decision trees AdaBoost Bagged trees k-nn Discriminant

17 挑战 : 使用图像进行机器学习的流程 Training Data Feature Extraction Learning or Modelling Challenge 1 Challenge 2 Challenge 3 Challenge 2 Classifier / Model hat Input Image Feature Extraction Classification 20

18 Common Challenges for Machine Learning with Images 挑战 1: 处理大量图像数据集挑战 2: 如何从图像中提取差别信息挑战 3: 如何使用机器学习技术对问题建模方便操作的大量图像数据集 imageset 将 Bag of words 用于特征提取更多特征提取函数见 Computer Vision System Toolbox Classification APP 21

问题建模方便操作的大量图像数据集 imageset 将 Bag of words 用于特征提取

19 Examples of Object Recognition/Classification Automatic scene categorization Biometrics Face recognition IRIS recognition Fingerprint recognition Part recognition for factory automation Autonomous robotics 22

recognition IRIS recognition Fingerprint

20 题纲基于 MATLAB 的人工智能设计流程介绍演示人工智能流程和常见挑战针对实时视频进行目标检测训练目标检测器并行计算加速集群 GPU/ 多 GPU 加速总结 26

21 基于 MATLAB 的目标检测多种方法 : 1. 特征检测提取和匹配 2. 预先训练好的模型 3. Out-of-the box frameworks 从头开始设计分类器两类分类器 ( classes are object and everything else ) 27

22 Object Detection is Classification Sliding Window Extract Feature Classify Patch Label, Score 28

23 级联分类器 (Cascade classifier ) 灵感 Boosting Cancer detection CEA X AFP X CA153 XCA199 X CA125 =+99% accuracy 级联分类器的结构 (Cascade classifier Structure) 29

24 典型应用 : Viola Jones Algorithm Face detection 效率 : will quit at 1 st stage for the majority of the area. Features: HoG overall shape of the object people,car detection LBP Haar Fine scale texture face,nose,eye etc Tip : To save time, use small image set and use HoG & LBP to decide whether it is good with fine scale texture or overall shape. 30

25 总结 : 创建目标检测器使用 MATLAB Apps 简化训练过程耗时采用目标检测器架构快速搭建自定制目标检测器大量预先训练的检测器可供使用 31

26 题纲基于 MATLAB 的人工智能设计流程介绍演示人工智能流程和常见挑战针对实时视频进行目标检测训练目标检测器并行计算加速集群 GPU/ 多 GPU 加速总结 34

27 并行计算助理人工智能 MATLAB 并行计算实现 : 串行到并行 Worker Worker Worker TOOLBOXES BLOCKSETS Worker Worker Worker Worker Worker 35

28 MATLAB 并行计算实现 : 串行到并行 Worker Worker Worker TOOLBOXES BLOCKSETS Worker Worker Worker Worker Worker 36

29 MATLAB 并行计算工具箱使用并行计算工具箱 Desktop Computer Parallel Computing Toolbox 使本地计算加速充分发挥 CPU 和 GPU 的各自计算能力无需另外的计算资源 37

30 易于扩展到集群, 网格, 和云桌面计算 Parallel Computing Toolbox 计算集群 MATLAB Distributed Computing Server Scheduler 38

31 易于使用 CPU 并行编程实现 : 少量代码修改工具箱内部函数支持上层编程结构 parfor, batch, distributed 便于控制底层编程结构 createjob, labsend 39

32 题纲基于 MATLAB 的人工智能设计流程介绍演示人工智能流程和常见挑战针对实时视频进行目标检测训练目标检测器并行计算加速集群 GPU/ 多 GPU 加速总结 40

33 GPU 加速卷积神经网络案例 41

34 多 GPU 加速 useparallel yes no 'usegpu yes no only What if GPU + CPU? 自动在大型神经网络训练时保留 Checkpoints By default, checkpoint saves occur at most once every 60 seconds. Save to.mat file, and load easily 42

35 题纲基于 MATLAB 的人工智能设计流程介绍演示人工智能流程和常见挑战针对实时视频进行目标检测训练目标检测器并行计算加速集群 GPU/ 多 GPU 加速总结 43

36 Thank You! 45

《C语言基础入门》课程教学大纲

《C语言基础入门》课程教学大纲 C 语言开发入门教程课程教学大纲课程编号 :201409210011 学分 :5 学分学时 :58 学时 ( 其中 : 讲课学时 :39 学时上机学时 :19 学时 ) 先修课程 : 计算机导论后续课程 :C++ 程序设计适用专业 : 信息及其计算机相关专业开课部门 : 计