For my phd qualifiers, I have two in-class exams and one take-home. The take-home is Computer Vision. This is quite an interesting exam. We are given a paper, which we read,analyze and scrutinize. Then we answer some questions aimed at providing a critique of the technique discussed in the paper and if possible compare it with other contemporary techniques. At the end of it all, you are required to deliver a working implementation all in 4 days’ time. I am trying to keep my programming skills sharp. For my new research project I am trying to detect bicycles in video sequences. For this I am looking at a solid contour extraction method that handles clutter well. I can’t believe I am looking at Snakes for an answer which incidentally is prescribed in the reading list for my phd qualifiers. This was initially published in 1987 and it still is amazingly relevant. Take a look : http://www.markschulze.net/snakes/.
I am currently trying to answer this question:In the event of two or more bicycles passing side-by-side, if the contour embraces the bicycles collectively, what sort of feature matching and description can I possibly provide so as to count them correctly. If you are reading this please feel free to leave comments.
