Transcript
Page 1: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Introduction to Computer Vision

Olac Fuentes

Computer Science Department

University of Texas at El Paso

El Paso TX USA

What is Computer Vision

Computer Vision is the process of extracting knowledge about the world from one or more digital images

Digital Images

are 2D arrays (matrices) of numbers

Digital ImagesColor Images are formed with three

2-D arrays representing the Red Green and Blue components of the image

Computer Vision ndash Main Tasks

bull Model generationbull Object Recognitionbull Object Detectionbull Tracking

Computer Vision ndash Object DetectionDetecting Faces

Computer Vision ndash Object DetectionDetecting Faces

Computer Vision ndash Object DetectionDetecting Pedestrians

Computer Vision ndash Object DetectionDetecting Cars

Computer Vision ndash Object DetectionHow to do it

Idea Use Machine LearningTraining

Training Set bull Positive examples are images of objects that belong to the class of

interestbull Negative examples are images of objects that donrsquot belong to that

classTrain classifier using the training set

DetectionGiven an image to analyze apply classifier to every subimage (there are lots of them so a low false positive rate is important)

Face Detection ndash Training Images

Efficient Object DetectionViola amp Jones 2005

Idea 1 Classifier Structure

Build a cascade classifiers

Where stage i is simpler (and faster) than stage i+1

Efficient Object DetectionViola amp Jones 2005

Idea 2 FeaturesUse a large number of very simple features

Efficient Object DetectionViola amp Jones 2005

Idea 3 Feature ComputationCompute the features very efficiently using the integral image

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scales

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Better ideaBuild a detector for a single scaleDuring detection scale the image

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as features

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as featuresUsed simple voting classifiers and Adaboost to build a cascade of classifiers

Efficient Object DetectionHistograms of Gradients (Dalal 2005)

Histograms of Gradients (Dalal 2005)Used histograms of oriented gradients as features

Used Support Vector Machine as classifierBest results to date

Object Recognition

Owl

Duck

Toucan

Egret

TrainingTesting

Object Recognition ndash Face Recognition

Eigenfaces are a set of standardized face ingredients derived from statistical analysis of many pictures of faces

First four eigenfaces from the ATampT database

Eigenfacesbull One persons face might be made up of 10 from face 1 24 from face 2 and so on

Very few eigenvector terms are needed to give a fair likeness of most peoples faces

Eigenfaces provide a means of applying data compression to faces for identification purposes

Eigenfacesbull Let E1En be the eigenfaces obtained from a face database

Let F1Fm be the images in our trainingtesting sets (For the training images we also know the personrsquos identity)

The attributes of Fi are given by the sum of the pixel by pixel products of Fi and E1En that is Fi is represented by n numbers [FiE1 FiE2 FiEn]

Using the attribute vectors and the class information we can now construct a classifier

Tracking

Continuous detection of objects of interest in video streams

Tracking

Continuous detection of objects of interest in video streams

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometry

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 2: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

What is Computer Vision

Computer Vision is the process of extracting knowledge about the world from one or more digital images

Digital Images

are 2D arrays (matrices) of numbers

Digital ImagesColor Images are formed with three

2-D arrays representing the Red Green and Blue components of the image

Computer Vision ndash Main Tasks

bull Model generationbull Object Recognitionbull Object Detectionbull Tracking

Computer Vision ndash Object DetectionDetecting Faces

Computer Vision ndash Object DetectionDetecting Faces

Computer Vision ndash Object DetectionDetecting Pedestrians

Computer Vision ndash Object DetectionDetecting Cars

Computer Vision ndash Object DetectionHow to do it

Idea Use Machine LearningTraining

Training Set bull Positive examples are images of objects that belong to the class of

interestbull Negative examples are images of objects that donrsquot belong to that

classTrain classifier using the training set

DetectionGiven an image to analyze apply classifier to every subimage (there are lots of them so a low false positive rate is important)

Face Detection ndash Training Images

Efficient Object DetectionViola amp Jones 2005

Idea 1 Classifier Structure

Build a cascade classifiers

Where stage i is simpler (and faster) than stage i+1

Efficient Object DetectionViola amp Jones 2005

Idea 2 FeaturesUse a large number of very simple features

Efficient Object DetectionViola amp Jones 2005

Idea 3 Feature ComputationCompute the features very efficiently using the integral image

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scales

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Better ideaBuild a detector for a single scaleDuring detection scale the image

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as features

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as featuresUsed simple voting classifiers and Adaboost to build a cascade of classifiers

Efficient Object DetectionHistograms of Gradients (Dalal 2005)

Histograms of Gradients (Dalal 2005)Used histograms of oriented gradients as features

Used Support Vector Machine as classifierBest results to date

Object Recognition

Owl

Duck

Toucan

Egret

TrainingTesting

Object Recognition ndash Face Recognition

Eigenfaces are a set of standardized face ingredients derived from statistical analysis of many pictures of faces

First four eigenfaces from the ATampT database

Eigenfacesbull One persons face might be made up of 10 from face 1 24 from face 2 and so on

Very few eigenvector terms are needed to give a fair likeness of most peoples faces

Eigenfaces provide a means of applying data compression to faces for identification purposes

Eigenfacesbull Let E1En be the eigenfaces obtained from a face database

Let F1Fm be the images in our trainingtesting sets (For the training images we also know the personrsquos identity)

The attributes of Fi are given by the sum of the pixel by pixel products of Fi and E1En that is Fi is represented by n numbers [FiE1 FiE2 FiEn]

Using the attribute vectors and the class information we can now construct a classifier

Tracking

Continuous detection of objects of interest in video streams

Tracking

Continuous detection of objects of interest in video streams

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometry

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 3: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Digital Images

are 2D arrays (matrices) of numbers

Digital ImagesColor Images are formed with three

2-D arrays representing the Red Green and Blue components of the image

Computer Vision ndash Main Tasks

bull Model generationbull Object Recognitionbull Object Detectionbull Tracking

Computer Vision ndash Object DetectionDetecting Faces

Computer Vision ndash Object DetectionDetecting Faces

Computer Vision ndash Object DetectionDetecting Pedestrians

Computer Vision ndash Object DetectionDetecting Cars

Computer Vision ndash Object DetectionHow to do it

Idea Use Machine LearningTraining

Training Set bull Positive examples are images of objects that belong to the class of

interestbull Negative examples are images of objects that donrsquot belong to that

classTrain classifier using the training set

DetectionGiven an image to analyze apply classifier to every subimage (there are lots of them so a low false positive rate is important)

Face Detection ndash Training Images

Efficient Object DetectionViola amp Jones 2005

Idea 1 Classifier Structure

Build a cascade classifiers

Where stage i is simpler (and faster) than stage i+1

Efficient Object DetectionViola amp Jones 2005

Idea 2 FeaturesUse a large number of very simple features

Efficient Object DetectionViola amp Jones 2005

Idea 3 Feature ComputationCompute the features very efficiently using the integral image

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scales

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Better ideaBuild a detector for a single scaleDuring detection scale the image

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as features

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as featuresUsed simple voting classifiers and Adaboost to build a cascade of classifiers

Efficient Object DetectionHistograms of Gradients (Dalal 2005)

Histograms of Gradients (Dalal 2005)Used histograms of oriented gradients as features

Used Support Vector Machine as classifierBest results to date

Object Recognition

Owl

Duck

Toucan

Egret

TrainingTesting

Object Recognition ndash Face Recognition

Eigenfaces are a set of standardized face ingredients derived from statistical analysis of many pictures of faces

First four eigenfaces from the ATampT database

Eigenfacesbull One persons face might be made up of 10 from face 1 24 from face 2 and so on

Very few eigenvector terms are needed to give a fair likeness of most peoples faces

Eigenfaces provide a means of applying data compression to faces for identification purposes

Eigenfacesbull Let E1En be the eigenfaces obtained from a face database

Let F1Fm be the images in our trainingtesting sets (For the training images we also know the personrsquos identity)

The attributes of Fi are given by the sum of the pixel by pixel products of Fi and E1En that is Fi is represented by n numbers [FiE1 FiE2 FiEn]

Using the attribute vectors and the class information we can now construct a classifier

Tracking

Continuous detection of objects of interest in video streams

Tracking

Continuous detection of objects of interest in video streams

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometry

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 4: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Digital ImagesColor Images are formed with three

2-D arrays representing the Red Green and Blue components of the image

Computer Vision ndash Main Tasks

bull Model generationbull Object Recognitionbull Object Detectionbull Tracking

Computer Vision ndash Object DetectionDetecting Faces

Computer Vision ndash Object DetectionDetecting Faces

Computer Vision ndash Object DetectionDetecting Pedestrians

Computer Vision ndash Object DetectionDetecting Cars

Computer Vision ndash Object DetectionHow to do it

Idea Use Machine LearningTraining

Training Set bull Positive examples are images of objects that belong to the class of

interestbull Negative examples are images of objects that donrsquot belong to that

classTrain classifier using the training set

DetectionGiven an image to analyze apply classifier to every subimage (there are lots of them so a low false positive rate is important)

Face Detection ndash Training Images

Efficient Object DetectionViola amp Jones 2005

Idea 1 Classifier Structure

Build a cascade classifiers

Where stage i is simpler (and faster) than stage i+1

Efficient Object DetectionViola amp Jones 2005

Idea 2 FeaturesUse a large number of very simple features

Efficient Object DetectionViola amp Jones 2005

Idea 3 Feature ComputationCompute the features very efficiently using the integral image

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scales

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Better ideaBuild a detector for a single scaleDuring detection scale the image

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as features

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as featuresUsed simple voting classifiers and Adaboost to build a cascade of classifiers

Efficient Object DetectionHistograms of Gradients (Dalal 2005)

Histograms of Gradients (Dalal 2005)Used histograms of oriented gradients as features

Used Support Vector Machine as classifierBest results to date

Object Recognition

Owl

Duck

Toucan

Egret

TrainingTesting

Object Recognition ndash Face Recognition

Eigenfaces are a set of standardized face ingredients derived from statistical analysis of many pictures of faces

First four eigenfaces from the ATampT database

Eigenfacesbull One persons face might be made up of 10 from face 1 24 from face 2 and so on

Very few eigenvector terms are needed to give a fair likeness of most peoples faces

Eigenfaces provide a means of applying data compression to faces for identification purposes

Eigenfacesbull Let E1En be the eigenfaces obtained from a face database

Let F1Fm be the images in our trainingtesting sets (For the training images we also know the personrsquos identity)

The attributes of Fi are given by the sum of the pixel by pixel products of Fi and E1En that is Fi is represented by n numbers [FiE1 FiE2 FiEn]

Using the attribute vectors and the class information we can now construct a classifier

Tracking

Continuous detection of objects of interest in video streams

Tracking

Continuous detection of objects of interest in video streams

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometry

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 5: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Computer Vision ndash Main Tasks

bull Model generationbull Object Recognitionbull Object Detectionbull Tracking

Computer Vision ndash Object DetectionDetecting Faces

Computer Vision ndash Object DetectionDetecting Faces

Computer Vision ndash Object DetectionDetecting Pedestrians

Computer Vision ndash Object DetectionDetecting Cars

Computer Vision ndash Object DetectionHow to do it

Idea Use Machine LearningTraining

Training Set bull Positive examples are images of objects that belong to the class of

interestbull Negative examples are images of objects that donrsquot belong to that

classTrain classifier using the training set

DetectionGiven an image to analyze apply classifier to every subimage (there are lots of them so a low false positive rate is important)

Face Detection ndash Training Images

Efficient Object DetectionViola amp Jones 2005

Idea 1 Classifier Structure

Build a cascade classifiers

Where stage i is simpler (and faster) than stage i+1

Efficient Object DetectionViola amp Jones 2005

Idea 2 FeaturesUse a large number of very simple features

Efficient Object DetectionViola amp Jones 2005

Idea 3 Feature ComputationCompute the features very efficiently using the integral image

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scales

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Better ideaBuild a detector for a single scaleDuring detection scale the image

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as features

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as featuresUsed simple voting classifiers and Adaboost to build a cascade of classifiers

Efficient Object DetectionHistograms of Gradients (Dalal 2005)

Histograms of Gradients (Dalal 2005)Used histograms of oriented gradients as features

Used Support Vector Machine as classifierBest results to date

Object Recognition

Owl

Duck

Toucan

Egret

TrainingTesting

Object Recognition ndash Face Recognition

Eigenfaces are a set of standardized face ingredients derived from statistical analysis of many pictures of faces

First four eigenfaces from the ATampT database

Eigenfacesbull One persons face might be made up of 10 from face 1 24 from face 2 and so on

Very few eigenvector terms are needed to give a fair likeness of most peoples faces

Eigenfaces provide a means of applying data compression to faces for identification purposes

Eigenfacesbull Let E1En be the eigenfaces obtained from a face database

Let F1Fm be the images in our trainingtesting sets (For the training images we also know the personrsquos identity)

The attributes of Fi are given by the sum of the pixel by pixel products of Fi and E1En that is Fi is represented by n numbers [FiE1 FiE2 FiEn]

Using the attribute vectors and the class information we can now construct a classifier

Tracking

Continuous detection of objects of interest in video streams

Tracking

Continuous detection of objects of interest in video streams

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometry

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 6: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Computer Vision ndash Object DetectionDetecting Faces

Computer Vision ndash Object DetectionDetecting Faces

Computer Vision ndash Object DetectionDetecting Pedestrians

Computer Vision ndash Object DetectionDetecting Cars

Computer Vision ndash Object DetectionHow to do it

Idea Use Machine LearningTraining

Training Set bull Positive examples are images of objects that belong to the class of

interestbull Negative examples are images of objects that donrsquot belong to that

classTrain classifier using the training set

DetectionGiven an image to analyze apply classifier to every subimage (there are lots of them so a low false positive rate is important)

Face Detection ndash Training Images

Efficient Object DetectionViola amp Jones 2005

Idea 1 Classifier Structure

Build a cascade classifiers

Where stage i is simpler (and faster) than stage i+1

Efficient Object DetectionViola amp Jones 2005

Idea 2 FeaturesUse a large number of very simple features

Efficient Object DetectionViola amp Jones 2005

Idea 3 Feature ComputationCompute the features very efficiently using the integral image

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scales

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Better ideaBuild a detector for a single scaleDuring detection scale the image

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as features

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as featuresUsed simple voting classifiers and Adaboost to build a cascade of classifiers

Efficient Object DetectionHistograms of Gradients (Dalal 2005)

Histograms of Gradients (Dalal 2005)Used histograms of oriented gradients as features

Used Support Vector Machine as classifierBest results to date

Object Recognition

Owl

Duck

Toucan

Egret

TrainingTesting

Object Recognition ndash Face Recognition

Eigenfaces are a set of standardized face ingredients derived from statistical analysis of many pictures of faces

First four eigenfaces from the ATampT database

Eigenfacesbull One persons face might be made up of 10 from face 1 24 from face 2 and so on

Very few eigenvector terms are needed to give a fair likeness of most peoples faces

Eigenfaces provide a means of applying data compression to faces for identification purposes

Eigenfacesbull Let E1En be the eigenfaces obtained from a face database

Let F1Fm be the images in our trainingtesting sets (For the training images we also know the personrsquos identity)

The attributes of Fi are given by the sum of the pixel by pixel products of Fi and E1En that is Fi is represented by n numbers [FiE1 FiE2 FiEn]

Using the attribute vectors and the class information we can now construct a classifier

Tracking

Continuous detection of objects of interest in video streams

Tracking

Continuous detection of objects of interest in video streams

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometry

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 7: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Computer Vision ndash Object DetectionDetecting Faces

Computer Vision ndash Object DetectionDetecting Pedestrians

Computer Vision ndash Object DetectionDetecting Cars

Computer Vision ndash Object DetectionHow to do it

Idea Use Machine LearningTraining

Training Set bull Positive examples are images of objects that belong to the class of

interestbull Negative examples are images of objects that donrsquot belong to that

classTrain classifier using the training set

DetectionGiven an image to analyze apply classifier to every subimage (there are lots of them so a low false positive rate is important)

Face Detection ndash Training Images

Efficient Object DetectionViola amp Jones 2005

Idea 1 Classifier Structure

Build a cascade classifiers

Where stage i is simpler (and faster) than stage i+1

Efficient Object DetectionViola amp Jones 2005

Idea 2 FeaturesUse a large number of very simple features

Efficient Object DetectionViola amp Jones 2005

Idea 3 Feature ComputationCompute the features very efficiently using the integral image

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scales

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Better ideaBuild a detector for a single scaleDuring detection scale the image

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as features

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as featuresUsed simple voting classifiers and Adaboost to build a cascade of classifiers

Efficient Object DetectionHistograms of Gradients (Dalal 2005)

Histograms of Gradients (Dalal 2005)Used histograms of oriented gradients as features

Used Support Vector Machine as classifierBest results to date

Object Recognition

Owl

Duck

Toucan

Egret

TrainingTesting

Object Recognition ndash Face Recognition

Eigenfaces are a set of standardized face ingredients derived from statistical analysis of many pictures of faces

First four eigenfaces from the ATampT database

Eigenfacesbull One persons face might be made up of 10 from face 1 24 from face 2 and so on

Very few eigenvector terms are needed to give a fair likeness of most peoples faces

Eigenfaces provide a means of applying data compression to faces for identification purposes

Eigenfacesbull Let E1En be the eigenfaces obtained from a face database

Let F1Fm be the images in our trainingtesting sets (For the training images we also know the personrsquos identity)

The attributes of Fi are given by the sum of the pixel by pixel products of Fi and E1En that is Fi is represented by n numbers [FiE1 FiE2 FiEn]

Using the attribute vectors and the class information we can now construct a classifier

Tracking

Continuous detection of objects of interest in video streams

Tracking

Continuous detection of objects of interest in video streams

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometry

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 8: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Computer Vision ndash Object DetectionDetecting Pedestrians

Computer Vision ndash Object DetectionDetecting Cars

Computer Vision ndash Object DetectionHow to do it

Idea Use Machine LearningTraining

Training Set bull Positive examples are images of objects that belong to the class of

interestbull Negative examples are images of objects that donrsquot belong to that

classTrain classifier using the training set

DetectionGiven an image to analyze apply classifier to every subimage (there are lots of them so a low false positive rate is important)

Face Detection ndash Training Images

Efficient Object DetectionViola amp Jones 2005

Idea 1 Classifier Structure

Build a cascade classifiers

Where stage i is simpler (and faster) than stage i+1

Efficient Object DetectionViola amp Jones 2005

Idea 2 FeaturesUse a large number of very simple features

Efficient Object DetectionViola amp Jones 2005

Idea 3 Feature ComputationCompute the features very efficiently using the integral image

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scales

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Better ideaBuild a detector for a single scaleDuring detection scale the image

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as features

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as featuresUsed simple voting classifiers and Adaboost to build a cascade of classifiers

Efficient Object DetectionHistograms of Gradients (Dalal 2005)

Histograms of Gradients (Dalal 2005)Used histograms of oriented gradients as features

Used Support Vector Machine as classifierBest results to date

Object Recognition

Owl

Duck

Toucan

Egret

TrainingTesting

Object Recognition ndash Face Recognition

Eigenfaces are a set of standardized face ingredients derived from statistical analysis of many pictures of faces

First four eigenfaces from the ATampT database

Eigenfacesbull One persons face might be made up of 10 from face 1 24 from face 2 and so on

Very few eigenvector terms are needed to give a fair likeness of most peoples faces

Eigenfaces provide a means of applying data compression to faces for identification purposes

Eigenfacesbull Let E1En be the eigenfaces obtained from a face database

Let F1Fm be the images in our trainingtesting sets (For the training images we also know the personrsquos identity)

The attributes of Fi are given by the sum of the pixel by pixel products of Fi and E1En that is Fi is represented by n numbers [FiE1 FiE2 FiEn]

Using the attribute vectors and the class information we can now construct a classifier

Tracking

Continuous detection of objects of interest in video streams

Tracking

Continuous detection of objects of interest in video streams

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometry

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 9: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Computer Vision ndash Object DetectionDetecting Cars

Computer Vision ndash Object DetectionHow to do it

Idea Use Machine LearningTraining

Training Set bull Positive examples are images of objects that belong to the class of

interestbull Negative examples are images of objects that donrsquot belong to that

classTrain classifier using the training set

DetectionGiven an image to analyze apply classifier to every subimage (there are lots of them so a low false positive rate is important)

Face Detection ndash Training Images

Efficient Object DetectionViola amp Jones 2005

Idea 1 Classifier Structure

Build a cascade classifiers

Where stage i is simpler (and faster) than stage i+1

Efficient Object DetectionViola amp Jones 2005

Idea 2 FeaturesUse a large number of very simple features

Efficient Object DetectionViola amp Jones 2005

Idea 3 Feature ComputationCompute the features very efficiently using the integral image

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scales

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Better ideaBuild a detector for a single scaleDuring detection scale the image

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as features

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as featuresUsed simple voting classifiers and Adaboost to build a cascade of classifiers

Efficient Object DetectionHistograms of Gradients (Dalal 2005)

Histograms of Gradients (Dalal 2005)Used histograms of oriented gradients as features

Used Support Vector Machine as classifierBest results to date

Object Recognition

Owl

Duck

Toucan

Egret

TrainingTesting

Object Recognition ndash Face Recognition

Eigenfaces are a set of standardized face ingredients derived from statistical analysis of many pictures of faces

First four eigenfaces from the ATampT database

Eigenfacesbull One persons face might be made up of 10 from face 1 24 from face 2 and so on

Very few eigenvector terms are needed to give a fair likeness of most peoples faces

Eigenfaces provide a means of applying data compression to faces for identification purposes

Eigenfacesbull Let E1En be the eigenfaces obtained from a face database

Let F1Fm be the images in our trainingtesting sets (For the training images we also know the personrsquos identity)

The attributes of Fi are given by the sum of the pixel by pixel products of Fi and E1En that is Fi is represented by n numbers [FiE1 FiE2 FiEn]

Using the attribute vectors and the class information we can now construct a classifier

Tracking

Continuous detection of objects of interest in video streams

Tracking

Continuous detection of objects of interest in video streams

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometry

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 10: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Computer Vision ndash Object DetectionHow to do it

Idea Use Machine LearningTraining

Training Set bull Positive examples are images of objects that belong to the class of

interestbull Negative examples are images of objects that donrsquot belong to that

classTrain classifier using the training set

DetectionGiven an image to analyze apply classifier to every subimage (there are lots of them so a low false positive rate is important)

Face Detection ndash Training Images

Efficient Object DetectionViola amp Jones 2005

Idea 1 Classifier Structure

Build a cascade classifiers

Where stage i is simpler (and faster) than stage i+1

Efficient Object DetectionViola amp Jones 2005

Idea 2 FeaturesUse a large number of very simple features

Efficient Object DetectionViola amp Jones 2005

Idea 3 Feature ComputationCompute the features very efficiently using the integral image

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scales

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Better ideaBuild a detector for a single scaleDuring detection scale the image

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as features

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as featuresUsed simple voting classifiers and Adaboost to build a cascade of classifiers

Efficient Object DetectionHistograms of Gradients (Dalal 2005)

Histograms of Gradients (Dalal 2005)Used histograms of oriented gradients as features

Used Support Vector Machine as classifierBest results to date

Object Recognition

Owl

Duck

Toucan

Egret

TrainingTesting

Object Recognition ndash Face Recognition

Eigenfaces are a set of standardized face ingredients derived from statistical analysis of many pictures of faces

First four eigenfaces from the ATampT database

Eigenfacesbull One persons face might be made up of 10 from face 1 24 from face 2 and so on

Very few eigenvector terms are needed to give a fair likeness of most peoples faces

Eigenfaces provide a means of applying data compression to faces for identification purposes

Eigenfacesbull Let E1En be the eigenfaces obtained from a face database

Let F1Fm be the images in our trainingtesting sets (For the training images we also know the personrsquos identity)

The attributes of Fi are given by the sum of the pixel by pixel products of Fi and E1En that is Fi is represented by n numbers [FiE1 FiE2 FiEn]

Using the attribute vectors and the class information we can now construct a classifier

Tracking

Continuous detection of objects of interest in video streams

Tracking

Continuous detection of objects of interest in video streams

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometry

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 11: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Face Detection ndash Training Images

Efficient Object DetectionViola amp Jones 2005

Idea 1 Classifier Structure

Build a cascade classifiers

Where stage i is simpler (and faster) than stage i+1

Efficient Object DetectionViola amp Jones 2005

Idea 2 FeaturesUse a large number of very simple features

Efficient Object DetectionViola amp Jones 2005

Idea 3 Feature ComputationCompute the features very efficiently using the integral image

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scales

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Better ideaBuild a detector for a single scaleDuring detection scale the image

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as features

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as featuresUsed simple voting classifiers and Adaboost to build a cascade of classifiers

Efficient Object DetectionHistograms of Gradients (Dalal 2005)

Histograms of Gradients (Dalal 2005)Used histograms of oriented gradients as features

Used Support Vector Machine as classifierBest results to date

Object Recognition

Owl

Duck

Toucan

Egret

TrainingTesting

Object Recognition ndash Face Recognition

Eigenfaces are a set of standardized face ingredients derived from statistical analysis of many pictures of faces

First four eigenfaces from the ATampT database

Eigenfacesbull One persons face might be made up of 10 from face 1 24 from face 2 and so on

Very few eigenvector terms are needed to give a fair likeness of most peoples faces

Eigenfaces provide a means of applying data compression to faces for identification purposes

Eigenfacesbull Let E1En be the eigenfaces obtained from a face database

Let F1Fm be the images in our trainingtesting sets (For the training images we also know the personrsquos identity)

The attributes of Fi are given by the sum of the pixel by pixel products of Fi and E1En that is Fi is represented by n numbers [FiE1 FiE2 FiEn]

Using the attribute vectors and the class information we can now construct a classifier

Tracking

Continuous detection of objects of interest in video streams

Tracking

Continuous detection of objects of interest in video streams

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometry

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 12: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Efficient Object DetectionViola amp Jones 2005

Idea 1 Classifier Structure

Build a cascade classifiers

Where stage i is simpler (and faster) than stage i+1

Efficient Object DetectionViola amp Jones 2005

Idea 2 FeaturesUse a large number of very simple features

Efficient Object DetectionViola amp Jones 2005

Idea 3 Feature ComputationCompute the features very efficiently using the integral image

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scales

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Better ideaBuild a detector for a single scaleDuring detection scale the image

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as features

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as featuresUsed simple voting classifiers and Adaboost to build a cascade of classifiers

Efficient Object DetectionHistograms of Gradients (Dalal 2005)

Histograms of Gradients (Dalal 2005)Used histograms of oriented gradients as features

Used Support Vector Machine as classifierBest results to date

Object Recognition

Owl

Duck

Toucan

Egret

TrainingTesting

Object Recognition ndash Face Recognition

Eigenfaces are a set of standardized face ingredients derived from statistical analysis of many pictures of faces

First four eigenfaces from the ATampT database

Eigenfacesbull One persons face might be made up of 10 from face 1 24 from face 2 and so on

Very few eigenvector terms are needed to give a fair likeness of most peoples faces

Eigenfaces provide a means of applying data compression to faces for identification purposes

Eigenfacesbull Let E1En be the eigenfaces obtained from a face database

Let F1Fm be the images in our trainingtesting sets (For the training images we also know the personrsquos identity)

The attributes of Fi are given by the sum of the pixel by pixel products of Fi and E1En that is Fi is represented by n numbers [FiE1 FiE2 FiEn]

Using the attribute vectors and the class information we can now construct a classifier

Tracking

Continuous detection of objects of interest in video streams

Tracking

Continuous detection of objects of interest in video streams

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometry

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 13: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Efficient Object DetectionViola amp Jones 2005

Idea 2 FeaturesUse a large number of very simple features

Efficient Object DetectionViola amp Jones 2005

Idea 3 Feature ComputationCompute the features very efficiently using the integral image

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scales

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Better ideaBuild a detector for a single scaleDuring detection scale the image

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as features

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as featuresUsed simple voting classifiers and Adaboost to build a cascade of classifiers

Efficient Object DetectionHistograms of Gradients (Dalal 2005)

Histograms of Gradients (Dalal 2005)Used histograms of oriented gradients as features

Used Support Vector Machine as classifierBest results to date

Object Recognition

Owl

Duck

Toucan

Egret

TrainingTesting

Object Recognition ndash Face Recognition

Eigenfaces are a set of standardized face ingredients derived from statistical analysis of many pictures of faces

First four eigenfaces from the ATampT database

Eigenfacesbull One persons face might be made up of 10 from face 1 24 from face 2 and so on

Very few eigenvector terms are needed to give a fair likeness of most peoples faces

Eigenfaces provide a means of applying data compression to faces for identification purposes

Eigenfacesbull Let E1En be the eigenfaces obtained from a face database

Let F1Fm be the images in our trainingtesting sets (For the training images we also know the personrsquos identity)

The attributes of Fi are given by the sum of the pixel by pixel products of Fi and E1En that is Fi is represented by n numbers [FiE1 FiE2 FiEn]

Using the attribute vectors and the class information we can now construct a classifier

Tracking

Continuous detection of objects of interest in video streams

Tracking

Continuous detection of objects of interest in video streams

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometry

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 14: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Efficient Object DetectionViola amp Jones 2005

Idea 3 Feature ComputationCompute the features very efficiently using the integral image

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scales

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Better ideaBuild a detector for a single scaleDuring detection scale the image

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as features

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as featuresUsed simple voting classifiers and Adaboost to build a cascade of classifiers

Efficient Object DetectionHistograms of Gradients (Dalal 2005)

Histograms of Gradients (Dalal 2005)Used histograms of oriented gradients as features

Used Support Vector Machine as classifierBest results to date

Object Recognition

Owl

Duck

Toucan

Egret

TrainingTesting

Object Recognition ndash Face Recognition

Eigenfaces are a set of standardized face ingredients derived from statistical analysis of many pictures of faces

First four eigenfaces from the ATampT database

Eigenfacesbull One persons face might be made up of 10 from face 1 24 from face 2 and so on

Very few eigenvector terms are needed to give a fair likeness of most peoples faces

Eigenfaces provide a means of applying data compression to faces for identification purposes

Eigenfacesbull Let E1En be the eigenfaces obtained from a face database

Let F1Fm be the images in our trainingtesting sets (For the training images we also know the personrsquos identity)

The attributes of Fi are given by the sum of the pixel by pixel products of Fi and E1En that is Fi is represented by n numbers [FiE1 FiE2 FiEn]

Using the attribute vectors and the class information we can now construct a classifier

Tracking

Continuous detection of objects of interest in video streams

Tracking

Continuous detection of objects of interest in video streams

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometry

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 15: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scales

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Better ideaBuild a detector for a single scaleDuring detection scale the image

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as features

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as featuresUsed simple voting classifiers and Adaboost to build a cascade of classifiers

Efficient Object DetectionHistograms of Gradients (Dalal 2005)

Histograms of Gradients (Dalal 2005)Used histograms of oriented gradients as features

Used Support Vector Machine as classifierBest results to date

Object Recognition

Owl

Duck

Toucan

Egret

TrainingTesting

Object Recognition ndash Face Recognition

Eigenfaces are a set of standardized face ingredients derived from statistical analysis of many pictures of faces

First four eigenfaces from the ATampT database

Eigenfacesbull One persons face might be made up of 10 from face 1 24 from face 2 and so on

Very few eigenvector terms are needed to give a fair likeness of most peoples faces

Eigenfaces provide a means of applying data compression to faces for identification purposes

Eigenfacesbull Let E1En be the eigenfaces obtained from a face database

Let F1Fm be the images in our trainingtesting sets (For the training images we also know the personrsquos identity)

The attributes of Fi are given by the sum of the pixel by pixel products of Fi and E1En that is Fi is represented by n numbers [FiE1 FiE2 FiEn]

Using the attribute vectors and the class information we can now construct a classifier

Tracking

Continuous detection of objects of interest in video streams

Tracking

Continuous detection of objects of interest in video streams

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometry

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 16: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Better ideaBuild a detector for a single scaleDuring detection scale the image

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as features

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as featuresUsed simple voting classifiers and Adaboost to build a cascade of classifiers

Efficient Object DetectionHistograms of Gradients (Dalal 2005)

Histograms of Gradients (Dalal 2005)Used histograms of oriented gradients as features

Used Support Vector Machine as classifierBest results to date

Object Recognition

Owl

Duck

Toucan

Egret

TrainingTesting

Object Recognition ndash Face Recognition

Eigenfaces are a set of standardized face ingredients derived from statistical analysis of many pictures of faces

First four eigenfaces from the ATampT database

Eigenfacesbull One persons face might be made up of 10 from face 1 24 from face 2 and so on

Very few eigenvector terms are needed to give a fair likeness of most peoples faces

Eigenfaces provide a means of applying data compression to faces for identification purposes

Eigenfacesbull Let E1En be the eigenfaces obtained from a face database

Let F1Fm be the images in our trainingtesting sets (For the training images we also know the personrsquos identity)

The attributes of Fi are given by the sum of the pixel by pixel products of Fi and E1En that is Fi is represented by n numbers [FiE1 FiE2 FiEn]

Using the attribute vectors and the class information we can now construct a classifier

Tracking

Continuous detection of objects of interest in video streams

Tracking

Continuous detection of objects of interest in video streams

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometry

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 17: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Better ideaBuild a detector for a single scaleDuring detection scale the image

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as features

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as featuresUsed simple voting classifiers and Adaboost to build a cascade of classifiers

Efficient Object DetectionHistograms of Gradients (Dalal 2005)

Histograms of Gradients (Dalal 2005)Used histograms of oriented gradients as features

Used Support Vector Machine as classifierBest results to date

Object Recognition

Owl

Duck

Toucan

Egret

TrainingTesting

Object Recognition ndash Face Recognition

Eigenfaces are a set of standardized face ingredients derived from statistical analysis of many pictures of faces

First four eigenfaces from the ATampT database

Eigenfacesbull One persons face might be made up of 10 from face 1 24 from face 2 and so on

Very few eigenvector terms are needed to give a fair likeness of most peoples faces

Eigenfaces provide a means of applying data compression to faces for identification purposes

Eigenfacesbull Let E1En be the eigenfaces obtained from a face database

Let F1Fm be the images in our trainingtesting sets (For the training images we also know the personrsquos identity)

The attributes of Fi are given by the sum of the pixel by pixel products of Fi and E1En that is Fi is represented by n numbers [FiE1 FiE2 FiEn]

Using the attribute vectors and the class information we can now construct a classifier

Tracking

Continuous detection of objects of interest in video streams

Tracking

Continuous detection of objects of interest in video streams

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometry

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 18: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Efficient Object DetectionViola amp Jones 2005

Idea 4 Dealing with multiple scalesObvious solution

Build a detector for each possible scale

Better ideaBuild a detector for a single scaleDuring detection scale the image

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as features

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as featuresUsed simple voting classifiers and Adaboost to build a cascade of classifiers

Efficient Object DetectionHistograms of Gradients (Dalal 2005)

Histograms of Gradients (Dalal 2005)Used histograms of oriented gradients as features

Used Support Vector Machine as classifierBest results to date

Object Recognition

Owl

Duck

Toucan

Egret

TrainingTesting

Object Recognition ndash Face Recognition

Eigenfaces are a set of standardized face ingredients derived from statistical analysis of many pictures of faces

First four eigenfaces from the ATampT database

Eigenfacesbull One persons face might be made up of 10 from face 1 24 from face 2 and so on

Very few eigenvector terms are needed to give a fair likeness of most peoples faces

Eigenfaces provide a means of applying data compression to faces for identification purposes

Eigenfacesbull Let E1En be the eigenfaces obtained from a face database

Let F1Fm be the images in our trainingtesting sets (For the training images we also know the personrsquos identity)

The attributes of Fi are given by the sum of the pixel by pixel products of Fi and E1En that is Fi is represented by n numbers [FiE1 FiE2 FiEn]

Using the attribute vectors and the class information we can now construct a classifier

Tracking

Continuous detection of objects of interest in video streams

Tracking

Continuous detection of objects of interest in video streams

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometry

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 19: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as features

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as featuresUsed simple voting classifiers and Adaboost to build a cascade of classifiers

Efficient Object DetectionHistograms of Gradients (Dalal 2005)

Histograms of Gradients (Dalal 2005)Used histograms of oriented gradients as features

Used Support Vector Machine as classifierBest results to date

Object Recognition

Owl

Duck

Toucan

Egret

TrainingTesting

Object Recognition ndash Face Recognition

Eigenfaces are a set of standardized face ingredients derived from statistical analysis of many pictures of faces

First four eigenfaces from the ATampT database

Eigenfacesbull One persons face might be made up of 10 from face 1 24 from face 2 and so on

Very few eigenvector terms are needed to give a fair likeness of most peoples faces

Eigenfaces provide a means of applying data compression to faces for identification purposes

Eigenfacesbull Let E1En be the eigenfaces obtained from a face database

Let F1Fm be the images in our trainingtesting sets (For the training images we also know the personrsquos identity)

The attributes of Fi are given by the sum of the pixel by pixel products of Fi and E1En that is Fi is represented by n numbers [FiE1 FiE2 FiEn]

Using the attribute vectors and the class information we can now construct a classifier

Tracking

Continuous detection of objects of interest in video streams

Tracking

Continuous detection of objects of interest in video streams

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometry

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 20: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Efficient Object DetectionThe Modified census transform (Froba and Ernst 2004)

Used local intensity descriptors as featuresUsed simple voting classifiers and Adaboost to build a cascade of classifiers

Efficient Object DetectionHistograms of Gradients (Dalal 2005)

Histograms of Gradients (Dalal 2005)Used histograms of oriented gradients as features

Used Support Vector Machine as classifierBest results to date

Object Recognition

Owl

Duck

Toucan

Egret

TrainingTesting

Object Recognition ndash Face Recognition

Eigenfaces are a set of standardized face ingredients derived from statistical analysis of many pictures of faces

First four eigenfaces from the ATampT database

Eigenfacesbull One persons face might be made up of 10 from face 1 24 from face 2 and so on

Very few eigenvector terms are needed to give a fair likeness of most peoples faces

Eigenfaces provide a means of applying data compression to faces for identification purposes

Eigenfacesbull Let E1En be the eigenfaces obtained from a face database

Let F1Fm be the images in our trainingtesting sets (For the training images we also know the personrsquos identity)

The attributes of Fi are given by the sum of the pixel by pixel products of Fi and E1En that is Fi is represented by n numbers [FiE1 FiE2 FiEn]

Using the attribute vectors and the class information we can now construct a classifier

Tracking

Continuous detection of objects of interest in video streams

Tracking

Continuous detection of objects of interest in video streams

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometry

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 21: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Efficient Object DetectionHistograms of Gradients (Dalal 2005)

Histograms of Gradients (Dalal 2005)Used histograms of oriented gradients as features

Used Support Vector Machine as classifierBest results to date

Object Recognition

Owl

Duck

Toucan

Egret

TrainingTesting

Object Recognition ndash Face Recognition

Eigenfaces are a set of standardized face ingredients derived from statistical analysis of many pictures of faces

First four eigenfaces from the ATampT database

Eigenfacesbull One persons face might be made up of 10 from face 1 24 from face 2 and so on

Very few eigenvector terms are needed to give a fair likeness of most peoples faces

Eigenfaces provide a means of applying data compression to faces for identification purposes

Eigenfacesbull Let E1En be the eigenfaces obtained from a face database

Let F1Fm be the images in our trainingtesting sets (For the training images we also know the personrsquos identity)

The attributes of Fi are given by the sum of the pixel by pixel products of Fi and E1En that is Fi is represented by n numbers [FiE1 FiE2 FiEn]

Using the attribute vectors and the class information we can now construct a classifier

Tracking

Continuous detection of objects of interest in video streams

Tracking

Continuous detection of objects of interest in video streams

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometry

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 22: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Object Recognition

Owl

Duck

Toucan

Egret

TrainingTesting

Object Recognition ndash Face Recognition

Eigenfaces are a set of standardized face ingredients derived from statistical analysis of many pictures of faces

First four eigenfaces from the ATampT database

Eigenfacesbull One persons face might be made up of 10 from face 1 24 from face 2 and so on

Very few eigenvector terms are needed to give a fair likeness of most peoples faces

Eigenfaces provide a means of applying data compression to faces for identification purposes

Eigenfacesbull Let E1En be the eigenfaces obtained from a face database

Let F1Fm be the images in our trainingtesting sets (For the training images we also know the personrsquos identity)

The attributes of Fi are given by the sum of the pixel by pixel products of Fi and E1En that is Fi is represented by n numbers [FiE1 FiE2 FiEn]

Using the attribute vectors and the class information we can now construct a classifier

Tracking

Continuous detection of objects of interest in video streams

Tracking

Continuous detection of objects of interest in video streams

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometry

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 23: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Object Recognition ndash Face Recognition

Eigenfaces are a set of standardized face ingredients derived from statistical analysis of many pictures of faces

First four eigenfaces from the ATampT database

Eigenfacesbull One persons face might be made up of 10 from face 1 24 from face 2 and so on

Very few eigenvector terms are needed to give a fair likeness of most peoples faces

Eigenfaces provide a means of applying data compression to faces for identification purposes

Eigenfacesbull Let E1En be the eigenfaces obtained from a face database

Let F1Fm be the images in our trainingtesting sets (For the training images we also know the personrsquos identity)

The attributes of Fi are given by the sum of the pixel by pixel products of Fi and E1En that is Fi is represented by n numbers [FiE1 FiE2 FiEn]

Using the attribute vectors and the class information we can now construct a classifier

Tracking

Continuous detection of objects of interest in video streams

Tracking

Continuous detection of objects of interest in video streams

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometry

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 24: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Eigenfacesbull One persons face might be made up of 10 from face 1 24 from face 2 and so on

Very few eigenvector terms are needed to give a fair likeness of most peoples faces

Eigenfaces provide a means of applying data compression to faces for identification purposes

Eigenfacesbull Let E1En be the eigenfaces obtained from a face database

Let F1Fm be the images in our trainingtesting sets (For the training images we also know the personrsquos identity)

The attributes of Fi are given by the sum of the pixel by pixel products of Fi and E1En that is Fi is represented by n numbers [FiE1 FiE2 FiEn]

Using the attribute vectors and the class information we can now construct a classifier

Tracking

Continuous detection of objects of interest in video streams

Tracking

Continuous detection of objects of interest in video streams

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometry

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 25: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Eigenfacesbull Let E1En be the eigenfaces obtained from a face database

Let F1Fm be the images in our trainingtesting sets (For the training images we also know the personrsquos identity)

The attributes of Fi are given by the sum of the pixel by pixel products of Fi and E1En that is Fi is represented by n numbers [FiE1 FiE2 FiEn]

Using the attribute vectors and the class information we can now construct a classifier

Tracking

Continuous detection of objects of interest in video streams

Tracking

Continuous detection of objects of interest in video streams

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometry

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 26: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Tracking

Continuous detection of objects of interest in video streams

Tracking

Continuous detection of objects of interest in video streams

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometry

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 27: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Tracking

Continuous detection of objects of interest in video streams

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometry

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 28: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometry

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 29: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

ReconstructionBuild a 3D models of world given 2D Images

Most-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake images

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 30: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

ReconstructionBuild a 3D models of world given 2D ImagesMost-common Approach Stereo VisionbullInspired by human 3D perceptionbullUse two cameras of known geometrybullTake imagesbullFind correspondencesbullReconstruct using correspondences and known geometry

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 31: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Reconstruction

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 32: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Reconstruction

Problems with Stereo VisionFinding matches reliably is difficultCalibration is difficultIt hard to deal with featureless areasComputationally expensive

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 33: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Reconstruction

Microsoft to the rescue

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 34: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

Reconstruction

Microsoft to the rescue

Seriously

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 35: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

ReconstructionMicrosoft Kinect

Reconstruction using active illumination

Project a known pattern of light at an invisible wavelength

Learn the appearance of that pattern at different distances

Fast and easy

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 36: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

ReconstructionMicrosoft Kinect

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37
Page 37: Introduction to Computer Vision Olac Fuentes Computer Science Department University of Texas at El Paso El Paso, TX, U.S.A

ReconstructionMicrosoft Kinect

  • Introduction to Computer Vision
  • What is Computer Vision
  • Digital Images
  • Slide 4
  • Computer Vision ndash Main Tasks
  • Computer Vision ndash Object Detection Detecting Faces
  • Slide 7
  • Computer Vision ndash Object Detection Detecting Pedestrians
  • Computer Vision ndash Object Detection Detecting Cars
  • Computer Vision ndash Object Detection How to do it
  • Face Detection ndash Training Images
  • Efficient Object Detection Viola amp Jones 2005
  • Slide 13
  • Slide 14
  • Slide 15
  • Slide 16
  • Slide 17
  • Slide 18
  • Efficient Object Detection The Modified census transform (Froba and Ernst 2004)
  • Slide 20
  • Efficient Object Detection Histograms of Gradients (Dalal 2005)
  • Object Recognition
  • Object Recognition ndash Face Recognition
  • Eigenfaces
  • Slide 25
  • Tracking
  • Slide 27
  • Reconstruction
  • Slide 29
  • Slide 30
  • Slide 31
  • Slide 32
  • Slide 33
  • Slide 34
  • Slide 35
  • Slide 36
  • Slide 37

Top Related