Final Project 236861

June 25, 2014

This June has witnessed a rapid development of situation woven by a sequence of events. That’s quite extraordinary for my usually laid-back lifestyle. In retrospect, this is clearly due to the “I’ll-do-this-and-that-after-draft-submission” mentality. Now in between the stages of research, with more time at my disposal, one of the outcomes is the long overdue completion of the final project for the course numerical geometry of images.

Regularly I only spent 2-3 hours every Thursday evening from March till May with some holiday on-and-offs, during which I, or rather Roy, finished the bulk of the project – the MSER part. Starting from June, I implemented the remaining 10%. And with some very dragged-on Q&A sessions due to my TA’s upgraded status to post doc, I finally got my grade yesterday. In March, some of my classmates already presented their final results of a few days of concentrated work. Although our coding load was allegedly much more than theirs, to my peace of mind the actual number of hours we spent on it is very similar to that of my classmates.

My project topic was changed to implementation of Sparse Modeling of Intrinsic Correspondences at the TA Yonathan’s suggestion. But the major part of the work is on the precondition of this method, MSER in deformable shapes. I was really interested after understanding how it works; at the same time I was worried about graphs and trees. Certainly I understood them conceptually, but it was hard to imagine how to put them into codes. But it was never a serious concern because of my top-programmer-class boyfriend. Every Thursday he came back and came to the school, we sat together in my office. I would explain how things work and what needs to be done. Then as I watched him coding, he would explain how to do a certain thing or why he did it that way. I still chuckle when I recall that Roy compulsively argued to correct my bad coding practices which too often occurred, and that during the break we ordered pizza or hamburgers. Sometimes he’d plan out the next things to do and write down the pseudo code for me to implement as a homework. This indeed is the most complicated programming I’ve been involved to day. Eventually by the time we joyfully witnessed the stable features detected on Homer and Armadillo, pointers and classes are no longer my enemies; trees and graphs are my frequent guests.


Linked to ProjectReport994201309mila.pdfKudos also go to my friend Alexandre, with whom I had meaningful conversations about the half-edge representation and whose triangulation infrastructure we used to process the discretized shape. And of course the very dedicated, patient, encouraging, busy and lively teaching assistant Yonathan. Whenever I encountered problems or uncertainties I turned to him. By emails, FaceTime and Monday meetings, nothing was unresolved in the end.

Leave a Reply

Your email address will not be published. Required fields are marked *