CSCE 689: Machine Learning Methods in Computer Vision (Fall 2017)

 

(Abbreviation: MACHINE LEARN COMP VIS)

 

Instructor

Dr. Zhangyang (Atlas) Wang

Webpage: http://www.atlaswang.com

 

Course Description

Discussion of established and new trends in optimization and statistical modeling approaches, as well as classical algorithms, for computer vision, pattern recognition and image processing, addressing both theoretical and application challenges.

 

Course Goal

The students will digest and practice their knowledge and skills by class discussion and course presentation, and obtain in-depth experience with a particular topic through a final project.

 

Evaluation Metrics

Grading will be based on three in-class quizzes (10% each), paper presentation and open discussion (25%), and one final project (45%) (proposal 5% + presentation 15% + code review 10% + report 15).  There will be no final exam. 

 

For the paper presentation and open discussion, a student presents a paper of his/her own choice, at either of the two scheduled slots (week 7 with sparsity/low-dimensionality theme, and week 10 with deep learning theme). No team work is allowed. To earn full 25% credits, the student has to both present the chosen paper (20%), and actively involve Q&A discussions when others present (5%). Extra credits (>25%) will be given to students who volunteer to present for both times (+5%).

 

For the final project, collaborations and teamwork are encouraged, but must be coordinated and approve by the instructor. Extra credits (>50%) will be given to:

-       One project to receive the Best Project Award, voted by all class members. (+5%)

-       Projects of interdisciplinary topics and novel application domains. (+3%)

-       Projects completed by only one individual student (a.k.a. lone hero bonus). (+3%)

For late submission, each additional late day will incur a 10% penalty.

 

Quiz

30%

Paper presentation and open discussion

25%+

Final project

45%+

 

(A "+" denotes an area in which extra credit can be earned.)

 

The grading policy is as follows:

90-100:

A

80-89:

B

70-79:

C

60-69:

D

<60:

F

 

Project

It's important that you work on a real computer vision project, or a real problem in some relevant domain (examples: multimedia signal processing like speech or text, medical image processing, social media, remote sensing, etc.), so that you earn first-hand experience how the computational models are bridged with the high complexity and uncertainty of the real world. You're free and encouraged to develop your project ideas, or you can consult the instructor for suggestion. The instructor is available to discuss and shape the project if you like. The scale of the project should be one semester long. By the end of the semester, you should submit your code and data for this project, write a project report of 8 pages at maximum using the standard CVPR paper template, and prepare a class presentation. The instructor will be happy to help you develop promising project ideas into a formal publication during or after the semester, if you wish so.

 

Prerequisite

Students should have taken the following courses or equivalent: Data Structure and Algorithms (CSCE 221), Linear Algebra (MATH 304 or MATH 323), Numerical Methods (MATH 417).

Coding experiences with Matlab (recommended), C/C++ or Python are assumed.

Previous knowledge of computer vision, machine learning or data mining will be helpful, but not necessary.

 

Reading Materials

This course does not follow any textbook closely. Among many recommended readings are:

1.     Introduction to Machine Learning, Ethem Alpaydin (2014), MIT Press. [Book home page (3rd edition)] [Book home page (2nd edition)] [Book home page (1st edition)]

2.     Pattern Recognition and Machine Learning, Christopher M. Bishop (2006). [A Bayesian view]

3.     The Elements of Statistical Learning, Jerome H. Friedman, Robert Tibshirani, and Trevor Hastie (2001), Springer. [Warning: not so elementary but quite insightful]

4.     Computer Vision: Algorithms and Applications, Richard Szeliski (2010), Springer.

5.     Sparse Coding and its Applications in Computer Vision, Wang et. al. (2015), WorldScientific.

6.     Convex Optimization, Stephen Boyd and Lieven Vandenberghe (2004), Cambridge University Press. [Their CVX toolbox is a great Matlab-based convex optimization tool for beginners]

7.     Distributed optimization and statistical learning via the alternating direction method of multipliers, Stephen Boyd et. al. (2011). [Dedicated reference for ADMM]

8.     Linear Algebra and its Applications, Gilbert Strang (1988). [For those who want to simply keep a concise reference for linear algebra, my best recommendation is The Matrix Cookbook]

Lecture notes (in PDF format) will be uploaded to the course webpage no less than 24 hours before each class. It is your responsibility to download and bring the notes to the class.

 

Attendance and Make-up Policies

Every student should attend the class, unless you have an accepted excuse. Please check student rule 7 http://student-rules.tamu.edu/rule07 for details.

 

Academic Integrity

Aggie Code of Honor: An Aggie does not lie, cheat or steal, or tolerate those who do. see: Honor Council Rules and Procedures

 

Americans with Disabilities Act (ADA) Statement

The Americans with Disabilities Act (ADA) is a federal anti-discrimination statute that provides comprehensive civil rights protection for persons with disabilities. Among other things, this legislation requires that all students with disabilities be guaranteed a learning environment that provides for reasonable accommodation of their disabilities. If you believe you have a disability requiring an accommodation, please contact Disability Services, currently located in the Disability Services building at the Student Services at White Creek complex on west campus or call 979-845-1637. For additional information, visit http://disability.tamu.edu.

 

Tentative schedule

 

Week 1

Introduction

 

Basics I: Linear Algebra

 

 

Week 2

Basics II: Vector Space and Optimization

 

Basics III: Probability Theory and Statistical Learning

 

 

Week 3

Linear Regression and Classification

 

 

Week 4

Support Vector Machine

 

 

Week 5

Sparse Learning I: Theory and Algorithms

 

Sparse Learning II: Models and Variants

 

 

Week 6

Sparse Learning III: Synthesis, Analysis, Transform, and Approximation

 

Low Dimensionality in High-Dimension Data

 

 

Week 7

Sparsity and Low Dimensionality

[Paper Presentation and Open Discussion 1]

 

Manifold Learning and Regularization

 

 

Week 8

Kernel Methods

 

Graphical Models

 

 

Week 9

Deep Learning I: Past and Present

 

Deep Learning II: Hot and Promising

 

 

Week 10

A Peep at the Future of Deep Learning

[Paper Presentation and Open Discussion 2]

 

Modelling Sequential Data

 

 

Week 11

Application Topics I: Image Restoration and Enhancement

 

Application Topics II: Image Recognition and Clustering

 

 

Week 12

Application Topics III: Objection Detection and Segmentation

 

Application Topics IV: Continuous and Discrete Feature Learning

 

 

Week 13

Application Topics V: Feature Selection and Information Fusion

 

Final Project Presentations [2-3 classes]

 

 

Week 14

Final Project Presentations

 

Final Project Presentations

 

(Some schedule flexibility is reserved for my out-of-town travels, and invited speakers)