Description

This course provides an introduction to computer vision and image processing by emphasizing the middle ground between image processing and artificial intelligence, using pattern recognition tools. Topics include image formation and representation, feature detection, motion estimation and tracking, object detection and recognition, and case studies of current computer vision research. We will briefly cover deep networks and discuss how these are playing a stronger role in vision. Students will receive a broad presentation of the field and will be expected to implement several computer vision algorithms throughout the semester.

General Information

Prerequisites:
Basic knowledge of probability, linear algebra, and vector calculus. MATLAB programming experience and previous exposure to image processing will be helpful, but not required.
Course Website
http://www.cubs.buffalo.edu/~inwogu/teaching/Coursepage573_fa14/
All announcements, course material, and related information will be communicated through the course website. We will use piazza specifically for discussions.
Textbook
There is no required text for this course. Lecture notes will be posted on the course web site a few days before the lecture. The following books are recommended though:
Computer Vision: A Modern Approach by David Forsyth and Jean Ponce (2nd ed.)
Computer Vision: Algorithms and Applications by Richard Szeliski (PDF available online)
Grading
Grading will be comprised of:
Weekly quizzes: 25%
Programming assignments: 50%
Exam: 25%
- Class participation will also be taken into account.

Announcements

CSE 473/573 final grades sheet have been posted!
12/22/14 8:22 PM

The teaching staff has posted the final grade sheets on the class homepage under Resources: cse473finalgrades.pdf and cse573finalgrades.pdf

You can view them on the course page in the grade sheets section at: https://piazza.com/buffalo/fall2014/cse473573/resources

Thank you all for a great semester. Tomorrow is the last day to make any unofficial changes to the final grades so please email or drop by my office if you have any outstanding issues/concerns/questions.  Otherwise happy holidays!

Yours,
Ifeoma Nwogu

cse573quizzesNexam.pdf has been added to class homepage under Resources
12/22/14 12:17 PM

The teaching staff has posted a new grade_sheets resource.
Title: cse573quizzesNexam.pdf
You can view it on the course page: https://piazza.com/buffalo/fall2014/cse473573/resources

Updated gradesheet have been added to class homepage under Resources
12/20/14 4:33 PM

The teaching staff has updated all the grade sheets on the resource tab, including homework 4. Please review to make sure your scores are up to date. We are working on a script to calculate final scores and final grades and will keep you up to date.

If you have any issues/questions/concerns with your scores, especially homework 4, please see me or Radha before the end of the day Monday 12/22 when we will submit the final grades to the college.

Thanks!
Ifeoma

p.s. pls feel free to send emails too if you have any questions or concerns

Winding down HW4 and submission
12/16/14 4:38 AM

Dear Students,

It's a day to the deadline and I would like to strongly advise folks to wind down on much of the programming activities. Although it is tempting to continue tweaking features in order to get better accuracy numbers, please focus now on finishing up the final report. Much of the grade will come from the quality of the report and presentations of results. Please take a minute to review http://www.cubs.buffalo.edu/~inwogu/teaching/Coursepage573_fa14/project.html

Also, it is very important that students do NOT submit any .mat files with the final reports. This could potentially take up a lot of the space on the cse servers. If you already submitted a file that contains .mat files, please resubmit a reduced size version to overwrite it. This is IMPORTANT.

Good luck with finishing up and thank you again for all the active participation in the course!

Ifeoma

Deadline for HW4 only extended to Wedneday 12/17 at 5pm
12/12/14 11:07 AM

Dear Students,

Since there has been so much going on  this week with other courses (exams, final presentations, etc.), and it seems like some students are only just starting their HW4, we will be extending the deadline to 12/17 at 5pm. Students who are able to finish on-time should please do so and submit their work, and start their holidays early. But others who need the extra time should take it.

I will not run the collection script until 5pm on Wednesday, but do not submit after that because it will not count. To be on the safe side, just submit something early on that day! This gives Radha and I only 2 days to grade all the work and tally up the final grades!!!

Ifeoma

p.s. to correct the test-ids bug, re-download homework4.zip from the resource tab

Mid-term exam and quiz grades uploaded
12/10/14 5:13 PM

The mid-term exam grades along with the grades from the eight quizzes taken over the course of the semester have been posted on been on the class homepage under Resources I the Grade sheets section. The files for 473 and 573 are cse473quizzesNexam.pdf and cse573quizzesNexam.pdf respectively.

Please review carefully and drop by my office - Davis 113M - if there are any issues or questions. I will also post the final grades for homeworks 1 -3 shortly. Thanks.

Ifeoma

Runtime issues
12/09/14 12:02 PM

Dear Students,

Due to the runtime problems everyone is having across board, I suggest that we reduce the size of the problem by starting with a significantly smaller dataset to train and test. Instead of using the almost 400,000 superpixels generated, you can train on a small subset of test_idx and test on the same size of train_idx.

Once you are satisfied with the features you have chosen as well as your classification results, then select a bigger training dataset. Lastly, a few images (maybe 10 or so) and use the superpixels from these images to test your results and these can be incorporated to your final report. The commands below might be useful for performing the training on a smaller set of about 1000 superpixels:

len = length(train_ids);

randlist = randperm(len);

Ftrain = Ftrain(randlist (1:1000),:);

Ctrain = Ctrain(randlist (1:1000),:);

This sequence of commands randomly selects 1000 superpixels and you can use this for initial training. Then maybe move up to 10,000 and if bold enough to 100,000 and finally to 300,000+. Record all your results and maybe plot how accuracies change with the number of training superpixels (optional).

A version of homework 4 files
12/05/14 1:44 PM

Hello All,

I just posted a cleaned up version of the homework instructions. Nothing has changed in the original instructions, but this is a little clearer to read. Also, I've uploaded some helper files for homework 4 on the resource page on piazza. Again today, we will go over the details in class along with some logistics for the course. Thanks everyone for a very enlightening semester.

Ifeoma

Staff Office Hours
NameOffice Hours
Ifeoma Nwogu
When?
Where?
Radhakrishna Dasari
When?
Where?
Devansh Arpit
When?
Where?

Lecture Notes