Homework 2: Face Up To It. Due March 4th
Overview
In this homework assignment you're going to demonstrate basic code level facility with face detection in OpenCV and video. This relatively short and quick assignment will serve to get you prepared for the second major semester assignment, which will involve face detection in PaSC handheld videos.
To being, you should review and run the Cascade Classifier Tutorial
Start with One Video
You are initially being provided with one handheld video, not stabilized, and you are to run the OpenCV cascade face detector on each frame of this video. You can obtain this video from the CS 510 unix account.
- /s/bach/b/class/cs510/spring2015/homework02/02463d3328.mp4
You have some latitude in how you display the results, but at a minimum you should be able to direct your code to place individual frames from the video in a subfolder with detected faces annoted visually by bounding boxes. You should also be able to direct your code to show results dynamically on the screen.
Some Possible Extras
The homework assignment as described thus far is worth 90 out of 100 points. The last 10 points are available for doing something extra. For example, running one or more additional videos and being ready to discuss the results. Another example, there are alternative trained cascade classifiers available and you could compare the standard against an alternative. You should feel free to email short descriptions of what you intend to do that is extra in advance to get an opinion about your choice.
Presenting your Running Video Face detector
There is no actual material that you need to submit in order to complete this homework. Instead, you will sit down one-on-one with the instructor and run your code. That may be your CS account on the CS department UNIX machines. You will be graded on how well your code performs the task, displays the result, and on your ability to explain components in your code. These presentations will be made on Wednesday, March 4th.
Addendum
As discussed in class, most students got the video reading working
smoothly, but there were problems using straight C++ on department
machines. I have now prepared a very simple test program, both C++
and Python, which work for me when I following the LD_LIBARARY_PATH
instructions in the README file. So, if you are still having
problems, retrieve the file
