CS 5501 (and CS4501) - "Big Data and Visual Search"

In Spring 2015, I am teaching "CS 5501 (and CS 4501) - big data and visual search" class. In this class, students will gain experience on processing large data sets. As a particular data application field, this class will cover mostly visual object recognition applications that utilize big data for learning and recognizing objects in images. Image retrieval, image matching, image parsing, object recognition are example applications that will be included in this class.

Time and location: Both CS5501 and CS4501 will take place on Tues.-Thurs., 5-6:15PM in Thornton E304.

Text Book: Kristen Grauman, Bastian Leibe, "Visual Object Recognition", Morgan & Claypool, Synthesis Lectures on Artificial Intelligence and Machine Learning, April 2011, 181 pages.

Prerequisites

Some previous exposure to "basics" of probability, matrix manupulations and optimization is helpful. Matlab will be the main programming and processing environment for this class. Optionally, you can also use OpenCV with C/C++.

For CS4501: CS2150, and a linear algebra class (such as UVA MATH 3350, MATH 3351, APMA 3080),
For CS5501: An introductory programming course (similar to UVA CS 1110 or CS 1111 or CS 1112) and a linear algebra class (such as UVA MATH 3350, MATH 3351, APMA 3080).

If you have questions about the prerequisites, please send an email to the instructor.

Tentative Syllabus

• Introduction,

• Visual search fundamentals: thresholding, edge detection, homography, histograms,

• Introduction to large image data sets (including PASCAL and ImageNet data sets),

• Features and attributes (SIFT, HOG, color histograms, Gabor Features),

• K-NN, Clustering and bag of words, Fisher vector, VLAD,

• RANSAC Algorithm,

• Classification including Kernel-machines (support vector machines, SVM, and multiple kernel learning, MKL,) and (sparse) Bayesian Learning,

• Introduction to Optimization (including constrained and unconstrained optimization),

• Various applications on PASCAL and ImageNet data sets including: Image retrieval, Image matching, Object recognition in image, Scene parsing, Semantic segmentation,

• Deep learning.

Grading

In this class, we will have assignments, midterms, presentations and projects. There is NO final exam (instead you will have a final project).
• Assignments: There are frequent (almost weekly) assignments in this class. Most assignments will use Matlab. (All the assignments will form the 30% of the final grade and the lowest 2 assignments will not count towards your final grade.
• Midterm: There is one midterm. That will be an hour (in-class) exam. (15% of the final grade.)
• Paper Presentation: Each student will present a paper. (15% of the final grade.)
• Final Project: There is no final exam in this class and instead, there will be a final project. The final project will be implemented on Matlab, and you will prepare a paper summarizing your findings, techniques and results. (40% of the final grade.)
Late submisisons will not be graded!

Lecture Notes

Lecture notes/slides will be available on Collab!