Teaching

Fall 2017: 776 Computer Vision

Slides

Introduction

Cameras

Distortion

Light, Color, and Extrinsics

Camera Geometry

Filters

Features

Features (cont.) and RANSAC

Robust Estimation (cont.)

Tracking and Optical Flow

Structure from Motion, Structure from Motion II

Intro to CNNs I, Intro to CNNs II

CNN Architectures

RNNs and GANs

Deep Learning for 3D Vision I, Deep Learning for 3D Vision II

Stereo Vision I, II, III

Assignments

  1. Assignment (due Sept. 6)
  2. Bayer Pattern / Distortion Estimation / Dolly Zoom (all three due Sept. 20)
  3. Feature Extraction (due Oct. 2)
  4. RANSAC (due Oct. 16)
  5. Triangulation / Bundle Adjustment (due Nov. 1)
  6. Project Proposals: Submit your PDF on Gradescope (due Oct. 30)

Spring 2017: 550 Algorithms and Analysis

Sakai Course Page (all information and materials posted on this page)

Fall 2016: 776 Computer Vision

There are two major research directions in computer vision: reconstruction aims at geometrically recreating the 3D world from the 2D photos/videos and recognition aims at extracting the semantics of objects and actions out of the images.

Prerequisites: Linear algebra, Matlab programing, image processing (useful but not required), basics of probabilities

Textbooks:

“Multiple View Geometry in Computer Vision”, Richard Hartley and Andrew Zisserman, Cambridge University Press, March 2004.

“Computer Vision: Algorithms and Applications” Richard Szeliski, Springer 2011, (online)

Grading: In this course we will significantly rely on practical programming exercises to gain a solid understanding of the discussed algorithms. Programming will be done in Matlab deploying the Image Processing and the Computer Vision Toolboxes or in python. Beyond the regular assignments there will be a larger final project. The active participation in class (attendance, asking and answering questions) will also be part of the grade. The weights assigned to different course components will be as follows:

Regular assignments: 45%

Final assignment: 25%

Participation: 30%

Syllabus:

I     Imaging

  • Camera models
  • Radiometric properties

II   Single View Characteristics

  • Filter
  • Features (corners, edges and blobs)

III  Two View Geometry

  • Homography
  • Epipolar geometry

IV. Robust Estimation

  • RANSAC
  • Hough transform

V.  Structure from motion and dense depth

  • Tracking
  • Optical flow
  • Camera calibration
  • Camera motion estimation
  • Two-view and multi-view stereo

VI. Recognition

  • Bags of features
  • Generative and discriminative models
  • Classification
  • Object recognition
  • Segmentation
  • Convolutional neural networks (CNN)

 Slides

Assignments

  1. Assignment
  2. Assignment
  3. Assignment
  4. Assignment

Fall 2015: 776 Computer Vision

There are two major research directions in computer vision: reconstruction aims at geometrically recreating the 3D world from the 2D photos/videos and recognition aims at extracting the semantics of objects and actions out of the images.

Prerequisites: Linear algebra, Matlab programing, image processing (useful but not required), basics of probabilities

Textbooks:

“Multiple View Geometry in Computer Vision”, Richard Hartley and Andrew Zisserman, Cambridge University Press, March 2004.

“Computer Vision: Algorithms and Applications” Richard Szeliski, Springer 2011, (online)

Grading: In this course we will significantly rely on practical programming exercises to gain a solid understanding of the discussed algorithms. Programming will be done in Matlab deploying the Image Processing and the Computer Vision Toolboxes. Beyond the regular assignments (3-4) there will be a larger final project. The active participation in class (attendance, asking and answering questions) will also be part of the grade. The weights assigned to different course components will be as follows:

Regular assignments: 45%

Final assignment: 25%

Participation: 30%

Syllabus:

I     Imaging

  • Camera models
  • Radiometric properties

II   Single View Characteristics

  • Filter
  • Features (corners, edges and blobs)

III  Two View Geometry

  • Homography
  • Epipolar geometry

IV. Robust Estimation

  • RANSAC
  • Hough transform

V.  Structure from motion and dense depth

  • Tracking
  • Optical flow
  • Camera calibration
  • Camera motion estimation
  • Two-view and multi-view stereo

VI. Recognition

  • Bags of features
  • Generative and discriminative models
  • Classification
  • Object recognition
  • Segmentation
  • Convolutional neural networks (CNN)

Slides:

Introduction

Camera I, Camera II

Light

Filters

Corners & Features I, II

Tracking & Model Fitting

Robust Estimation in Computer Vision

Structure from Motion

Dense Scene Geometry Estimation

Volumetric Reconstruction

Introduction to recognition

Localization

Learning

Face detection

Attributes

Deformable parts models

Segmentation

Convolutional neural networks

Transfer learning

Assignments:

1. Assignment

2. Assignment

3. Assignment

Final project

Fall 2014: 776 Computer Vision

There are two major research directions in computer vision: reconstruction aims at geometrically recreating the 3D world from the 2D photos/videos and recognition aims at extracting the semantics of objects and actions out of the images.

Prerequisites: Linear algebra, Matlab programing, image processing (useful but not required), basics of probabilities

Textbooks:

“Multiple View Geometry in Computer Vision”, Richard Hartley and Andrew Zisserman, Cambridge University Press, March 2004.

“Computer Vision: Algorithms and Applications” Richard Szeliski, Springer 2011, (online)

Grading: In this course we will significantly rely on practical programming exercises to gain a solid understanding of the discussed algorithms. Programming will be done in Matlab deploying the Image Processing and the Computer Vision Toolboxes. Beyond the regular assignments (3-4) there will be a larger final project. The active participation in class (attendance, asking and answering questions) will also be part of the grade. The weights assigned to different course components will be as follows:

Regular assignments: 45%

Final assignment: 25%

Participation: 30%

Syllabus:

I     Imaging

  • Camera models
  • Radiometric properties

II   Single View Characteristics

  • Filter
  • Features (corners, edges and blobs)

III  Two View Geometry

  • Homography
  • Epipolar geometry

IV. Robust Estimation

  • RANSAC
  • Hough transform

V.  Structure from motion and dense depth

  • Tracking
  • Optical flow
  • Camera calibration
  • Camera motion estimation
  • Two-view and multi-view stereo

VI. Recognition

  • Bags of features
  • Generative and discriminative models
  • Classification
  • Object recognition
  • Segmentation

Slides:

Introduction

Cameras

Light, Light part II

Filter

Corners part I, part II

Tracking

Fitting

Structure from Motion

Stereo

Dynamic Reconstruction I, II

Introduction Recognition

Localization

Face Detection I, II

Atributes I, II

Segmentation

Assignments:

1. Assignment

2. Assignment

3. Assignment

Final Project

——————————————-

Spring 2014  875 Recent advances in geometric computer vision and recognition

Introduction

Topic 1: Image based localization

Basic search concepts

Structure from Motion

Spring 2013: 776 Computer Vision

Feb. 3 – Feb. 17: Student presentations

There are two major research directions in computer vision: reconstruction aims at geometrically recreating the 3D world from the 2D photos/videos and recognition aims at extracting the semantics of objects and actions out of the images.

Prerequisites: Linear algebra, Matlab programing, image processing (useful but not required), basics of probabilities

Textbooks: 

“Multiple View Geometry in Computer Vision”, Richard Hartley and Andrew Zisserman, Cambridge University Press, March 2004.

“Computer Vision: Algorithms and Applications” Richard Szeliski, Springer 2011, (online)

Grading: In this course we will significantly rely on practical programming exercises to gain a solid understanding of the discussed algorithms. Programming will be done in Matlab deploying the Image Processing and the Computer Vision Toolboxes. Beyond the regular assignments (3-4) there will be a larger final project. The active participation in class (attendance, asking and answering questions) will also be part of the grade. The weights assigned to different course components will be as follows:

Regular assignments: 45%

Final assignment: 25%

Participation: 30%

Syllabus:

Geometry

  • Imaging
  • Single View Characteristics
  • Two View Geometry
  • Robust Estimation
  • Structure from motion and dense depth

Recognition

  • Classification
  • Face detection
  • Segmentation
  • Attributes

Slides:

January 10, Introduction

January 15, Camera Model I

January 17, Camera Model IIAssignment

January 22, Light I

January 24, Light II, Linear Filters

January 29, Filters

January 31, Corners

February 5, Tracking and Fitting

February 7, RANSAC, RECON, Hough TransformationAssignment

slides for classes from Feb 14-April 4 by Enrique Dunn

Fall 2012: COMP 066 Random Thoughts

Introduction [pptx]

Introduction to Excel [pptx]

Frequent disturbances in estimating probabilities [pptx]

Binomial distribution in real live [pptx]

Description of random variables [pptx]

Distributions [pptx]

Chance Events [pptx]

Polls (1) [pptx]

Polls (2) [pptx]

Bayes (1) [pptx]

Bayes (2) [pptx]

Final Project Instructions [pptx], Example projects [pptx]

Simpson Paradox [pptx]

How to predict election results [pptx]

Applications to security and privacy [pdf]

Hypothesis testing [pptx]

Hypothesis testing and Random Number [pptx]

Monte Carlo Simulation, RANSAC [pptx], Monte Carlo Simulation Excel [xlsx]

 

Spring 2012: 776 Computer Vision

There are two major research directions in computer vision: reconstruction aims at geometrically recreating the 3D world from the 2D photos/videos and recognition aims at extracting the semantics of objects and actions out of the images.

Prerequisites: Linear algebra, Matlab programing, image processing (useful but not required), basics of probabilities

Textbooks:

“Multiple View Geometry in Computer Vision”, Richard Hartley and Andrew Zisserman, Cambridge University Press, March 2004.

“Computer Vision: Algorithms and Applications” Richard Szeliski, Springer 2011, (online)

Grading: In this course we will significantly rely on practical programming exercises to gain a solid understanding of the discussed algorithms. Programming will be done in Matlab deploying the Image Processing and the Computer Vision Toolboxes. Beyond the regular assignments (3-4) there will be a larger final project. The active participation in class (attendance, asking and answering questions) will also be part of the grade. The weights assigned to different course components will be as follows:

Regular assignments: 45%

Final assignment: 25%

Participation: 30%

Syllabus:

I     Imaging

  • Camera models
  • Radiometric properties

II   Single View Characteristics

  • Filter
  • Features (corners, edges and blobs)

III  Two View Geometry

  • Homography
  • Epipolar geometry

IV. Robust Estimation

  • RANSAC
  • Hough transform

V.  Structure from motion and dense depth

  • Tracking
  • Optical flow
  • Camera calibration
  • Camera motion estimation
  • Two-view and multi-view stereo

VI. Recognition

  • Bags of features
  • Generative and discriminative models
  • Classification
  • Object recognition
  • Segmentation

Schedule

Introduction (ppt)

Cameras (ppt)

Light (ppt)

Filter (ppt)

Feature detection (ppt)

Fitting (ppt)

Fitting II (ppt)

Tracking (ppt)

Alignment (ppt)

Image Search (ppt)

Stereo (ppt)

Multi-view Stereo (ppt)

Structure from Motion (ppt)

Structure from Motion for Photo Collections (ppt)

Recognition Introduction (ppt)

Machine Learning (ppt)

Segmentation (ppt)

Segmentation & Motion Segmentation (ppt)

Attributes for Recognition (ppt)

Conclusions & Open Questions (ppt)

Feb. 2: 1 assignment (pdf)

March 1: 2nd assignment (pdf)

Final assignment (pdf)

Fall 2011: 875 Recent advances in geometric computer vision and recognition

August 29, Introduction

August 31, Geometry