斯坦福大学计算机视觉课件1
合集下载
相关主题
- 1、下载文档前请自行甄别文档内容的完整性，平台不提供额外的编辑、内容补充、找答案等附加服务。
- 2、"仅部分预览"的文档,不可在线预览部分如存在完整性等问题,可反馈申请退款(可完整预览的文档不适用该条件!)。
- 3、如文档侵犯您的权益，请联系客服反馈,我们会尽快为您处理(人工客服工作时间：9:00-18:30)。
9.913 Pattern Hale Waihona Puke Baiduecognition for Vision
Class I - Overview
Instructors: B. Heisele, Y. Ivanov, T. Poggio
Fall 2004
TOC
Administrivia Problems of Computer Vision and Pattern Recognition Overview of classes Quick review of Matlab
– Rods – Cones
Color is perceived by 3 types of cones
Density in Thousands per Square mm
– "Red" (64%) – "Green" (34%) – "Blue" (2%)
200 175 150 125 100 75 50 25 ROD DENSITY ROD DENSITY CONE DENSITY
All Visible Colors (CIE 1931 Chromaticity diagram):
Figure removed due to copyright reasons. Please see: http://www.cie.co.at/cie/ and http://www.ledtronics.com/datasheets/Pages/chromaticity/097b.htm
Images: A. Zisserman
Fall 2004
Images by A. Zisserman and P. Beardsley. Used with permission.
Pattern Recognition for Vision
Computer Vision - Object Detection \ Recognition
Fall 2004
Pattern Recognition for Vision
Classification
Test data Decision boundary Class model
?
Oranges
Apples
Training data
Images by MIT OCW.
Fall 2004 Pattern Recognition for Vision
Density Estimation
Image by MIT OCW.
, ∑ Generating density
Individual samples
Fall 2004
Pattern Recognition for Vision
Clustering
Data
Spatial clusters
Images by MIT OCW.
Determine objects' positions over multiple frames
Camera view
Connected components
Trajectories over time
Tracker – Chris Stauffer, Eric Grimson
An object
Fall 2004
Pattern Recognition for Vision
Machine Vision - Applications
Automatic quality control Robotics Perceptual interfaces – human/machine interactions Surveillance …
120,000,000 rods
6,000,000 cones
-80
-60
-40
-20
0
20
40
60
80
Angular Separation from Fovea (Degrees)
Figure by MIT OCW.
Fall 2004
Pattern Recognition for Vision
Color – CIE Chromaticity Diagram
Images removed due to copyright considerations.
Fall 2004
Pattern Recognition for Vision
Computer Vision – Shape-from-Shading
Find 3D surface parameters from a single image based on shading
Fall 2004 Pattern Recognition for Vision
Regression
Regressor Measurements
Jun 1 2 3 4 … Jul 1 2 3 … Aug 1 2 3 … Sep 1 2 3
Images by MIT OCW.
Fall 2004 Pattern Recognition for Vision
Imager:
Color properties Distortion Focal length F-stop (how wide the iris) Orientation
Surface properties:
Albedo (fraction of light reflected), Shape, Smoothness, Orientation
Fall 2004
Representation:
Bit depth Color space
Pattern Recognition for Vision
Vision - Color
Light is perceived by two types of receptors in the human eye
Images removed due to copyright considerations.
Images: NASA
Fall 2004 Pattern Recognition for Vision
Computer Vision - Stereo
Reconstruct 3D surface from 2 2D images taken simultaneously
Figures and photographs from: Stauffer, and Grimson. "Learning patterns of activity using real-time tracking." IEEE Transactions on Pattern Analysis and Machine Intelligence 22, no. 8 (August 2000): 747-757. Courtesy of IEEE, Chris Stauffer, and Eric Grimson. Copyright 2000 IEEE. Used with Permission.
Fall 2004 Pattern Recognition for Vision
Pattern Recognition
"Big Four" Problems of Machine Learning Classification Density Estimation Clustering Regression
Vision: Forsyth, Ponce, "Computer Vision: a Modern Approach" Machine Learning: Hastie, Tibshirani, Friedman, "The Elements of Statistical Learning"
Slides, links Papers, notes, tutorials Office hours
Vision – Image Formation and Processing
Topics Image formation Color spaces Point operators Neighborhood operators Edge detection Motion
Fall 2004
– Small weekly – Paper presentation/discussion – Final project
Fall 2004
Pattern Recognition for Vision
Syllabus
Sep 9 - Overview, Introduction Sep 16 - Vision - Image formation and processing Sep 23 - Vision – Feature extraction I Sep 30 - PR/Vis - Feature Extraction II/Bayesian decisions Oct 7 - PR - Density estimation Oct 14 - PR – Clasification Oct 21 - Biological Object Recognition Oct 28 - PR - Clustering Nov 4 - Paper Discussion Nov 11 - App I - Object Detection/Recognition Nov 18 - App II - Morphable models Nov 25 - No class - Thanksgiving day Dec 2 - App III - Tracking Dec 9 - App IV - Gesture and Action Recognition Dec 16 - Project presentation
Fall 2004
Y&B Y B B&Y Y papers B T Y&B proj All B T&B C&Y Y All
Pattern Recognition for Vision
Course Materials
Books
– Duda, Hart and Stork, Pattern Recognition – Optional - Mallot, Computational Vision: Information Processing in Perception and Visual Behavior – Suggested further reading
Pattern Recognition for Vision
Vision - Fundamentals
Image formation l l νl νe e e
Source properties:
Type of source Occlusions / shadows Direction
– No set hours – e-mail, or call
Fall 2004
Pattern Recognition for Vision
Computer Vision
Problems of Computer Vision Shape from Shading Stereo Structure from Motion Tracking Object Detection/Recognition Activity Detection/Recognition … many more….
Find objects in the image, determine what they are
Eg: Face detection and recognition:
Photograph by MIT OCW.
Fall 2004
Pattern Recognition for Vision
Computer Vision - Tracking
Fall 2004
Pattern Recognition for Vision
Administrivia
Instructors:
– Bernd Heisele – Yuri Ivanov – Tomaso Poggio
Meet
– Th 10-12
Credits: 10H Assignments:
Class I - Overview
Instructors: B. Heisele, Y. Ivanov, T. Poggio
Fall 2004
TOC
Administrivia Problems of Computer Vision and Pattern Recognition Overview of classes Quick review of Matlab
– Rods – Cones
Color is perceived by 3 types of cones
Density in Thousands per Square mm
– "Red" (64%) – "Green" (34%) – "Blue" (2%)
200 175 150 125 100 75 50 25 ROD DENSITY ROD DENSITY CONE DENSITY
All Visible Colors (CIE 1931 Chromaticity diagram):
Figure removed due to copyright reasons. Please see: http://www.cie.co.at/cie/ and http://www.ledtronics.com/datasheets/Pages/chromaticity/097b.htm
Images: A. Zisserman
Fall 2004
Images by A. Zisserman and P. Beardsley. Used with permission.
Pattern Recognition for Vision
Computer Vision - Object Detection \ Recognition
Fall 2004
Pattern Recognition for Vision
Classification
Test data Decision boundary Class model
?
Oranges
Apples
Training data
Images by MIT OCW.
Fall 2004 Pattern Recognition for Vision
Density Estimation
Image by MIT OCW.
, ∑ Generating density
Individual samples
Fall 2004
Pattern Recognition for Vision
Clustering
Data
Spatial clusters
Images by MIT OCW.
Determine objects' positions over multiple frames
Camera view
Connected components
Trajectories over time
Tracker – Chris Stauffer, Eric Grimson
An object
Fall 2004
Pattern Recognition for Vision
Machine Vision - Applications
Automatic quality control Robotics Perceptual interfaces – human/machine interactions Surveillance …
120,000,000 rods
6,000,000 cones
-80
-60
-40
-20
0
20
40
60
80
Angular Separation from Fovea (Degrees)
Figure by MIT OCW.
Fall 2004
Pattern Recognition for Vision
Color – CIE Chromaticity Diagram
Images removed due to copyright considerations.
Fall 2004
Pattern Recognition for Vision
Computer Vision – Shape-from-Shading
Find 3D surface parameters from a single image based on shading
Fall 2004 Pattern Recognition for Vision
Regression
Regressor Measurements
Jun 1 2 3 4 … Jul 1 2 3 … Aug 1 2 3 … Sep 1 2 3
Images by MIT OCW.
Fall 2004 Pattern Recognition for Vision
Imager:
Color properties Distortion Focal length F-stop (how wide the iris) Orientation
Surface properties:
Albedo (fraction of light reflected), Shape, Smoothness, Orientation
Fall 2004
Representation:
Bit depth Color space
Pattern Recognition for Vision
Vision - Color
Light is perceived by two types of receptors in the human eye
Images removed due to copyright considerations.
Images: NASA
Fall 2004 Pattern Recognition for Vision
Computer Vision - Stereo
Reconstruct 3D surface from 2 2D images taken simultaneously
Figures and photographs from: Stauffer, and Grimson. "Learning patterns of activity using real-time tracking." IEEE Transactions on Pattern Analysis and Machine Intelligence 22, no. 8 (August 2000): 747-757. Courtesy of IEEE, Chris Stauffer, and Eric Grimson. Copyright 2000 IEEE. Used with Permission.
Fall 2004 Pattern Recognition for Vision
Pattern Recognition
"Big Four" Problems of Machine Learning Classification Density Estimation Clustering Regression
Vision: Forsyth, Ponce, "Computer Vision: a Modern Approach" Machine Learning: Hastie, Tibshirani, Friedman, "The Elements of Statistical Learning"
Slides, links Papers, notes, tutorials Office hours
Vision – Image Formation and Processing
Topics Image formation Color spaces Point operators Neighborhood operators Edge detection Motion
Fall 2004
– Small weekly – Paper presentation/discussion – Final project
Fall 2004
Pattern Recognition for Vision
Syllabus
Sep 9 - Overview, Introduction Sep 16 - Vision - Image formation and processing Sep 23 - Vision – Feature extraction I Sep 30 - PR/Vis - Feature Extraction II/Bayesian decisions Oct 7 - PR - Density estimation Oct 14 - PR – Clasification Oct 21 - Biological Object Recognition Oct 28 - PR - Clustering Nov 4 - Paper Discussion Nov 11 - App I - Object Detection/Recognition Nov 18 - App II - Morphable models Nov 25 - No class - Thanksgiving day Dec 2 - App III - Tracking Dec 9 - App IV - Gesture and Action Recognition Dec 16 - Project presentation
Fall 2004
Y&B Y B B&Y Y papers B T Y&B proj All B T&B C&Y Y All
Pattern Recognition for Vision
Course Materials
Books
– Duda, Hart and Stork, Pattern Recognition – Optional - Mallot, Computational Vision: Information Processing in Perception and Visual Behavior – Suggested further reading
Pattern Recognition for Vision
Vision - Fundamentals
Image formation l l νl νe e e
Source properties:
Type of source Occlusions / shadows Direction
– No set hours – e-mail, or call
Fall 2004
Pattern Recognition for Vision
Computer Vision
Problems of Computer Vision Shape from Shading Stereo Structure from Motion Tracking Object Detection/Recognition Activity Detection/Recognition … many more….
Find objects in the image, determine what they are
Eg: Face detection and recognition:
Photograph by MIT OCW.
Fall 2004
Pattern Recognition for Vision
Computer Vision - Tracking
Fall 2004
Pattern Recognition for Vision
Administrivia
Instructors:
– Bernd Heisele – Yuri Ivanov – Tomaso Poggio
Meet
– Th 10-12
Credits: 10H Assignments: