EE 702 - COMPUTER VISION

Academic Year: 2014-15

Semester: Spring

Professor: Subhasis Chaudhuri

Prerequisites: Digital Image Processing (DIP).

The professor was upset with people having no background and the ones doing DIP in parallel with this course. It is highly advisable to have some knowledge about image processing.

Course contents:

This course basically deals with extracting 3D information on objects from insufficient data. (Yes, that’s all). Total course can be divided into 6 major parts:

  1. Depth from stereo
  2. Shape from stereo
  3. Shape from shading
  4. Optical flow
  5. Depth from defocus
  6. Hyperspectrum

Lectures:

Prof. SC extensively uses board to teach and doesn’t provide any slides. Course is majorly theory based (Yes, you are only supposed to do some 3 lab assignments). Attending every class is advisable as missing even a single class might make the rest of that chapter look gibberish. Also, problems from the chapters have numerous solutions with varied complexions, so you majorly have to rely on class notes to see which solutions were discussed.

You have to compulsorily submit every assignment which he asks to (told in advance by TAs), failing which might result in FR.

Assignments:

You will be asked to implement the algorithms taught in class in any software/IDE, using any libraries you like. You have to submit at least 2/3 out of 5/6 assignments. These are not easy and you will not be given any skeleton codes for your assistance. Most of the solutions involve direct implementation of some research papers.

Evaluation:

Assignment = 10+10 = 20% Paper Review = 10% you have to write short review/critic of some recent research papers. Midsem = 25% (Midsem is take-home, spanned over weekend) Question Paper Setting = 10% You have to set intriguing new questions ‘you’ encounter related to Computer Vision. Endsem = 35% Midsem and endsem papers are graded by Prof. himself.

Reviewed by - Nitin Chavan nitinbc@iitb.ac.in