When I came to U of T, I was under the impression that all I have to do is three quick projects. So I started working really hard, didn't sleep, changed project directions like socks, but nothing really came out for two years. Everything I tried either didn't work, was already done, or my supervisor convinced me that it was not interesting. Then I started to converge towards the problem of 3D surface reconstruction from single images. In fact, I was shocked how come I never heard about it before. The problem that seems to be the key to vision was never emphasized in vision courses or textbooks, because they focus mainly on what is known rather than what is open. We almost submitted a paper on planar curves, when we discovered the idea had been investigated more than 20 years before. Later I've seen this occurring to many students. In a world with more than 1000 theses in computer science and electrical engineering each year, it is only natural for similar people using the same sources to come up with similar ideas. In the past it was easier to reinvent ideas and publish them, but the internet completely changed the game.
At that point I concluded that I just didn't know computer vision well enough. I figured that in this business, it is better to make a small step in the right direction than a big step in the wrong direction. My conclusion was, and that was the hardest part, that in a world where everybody is rushing, the way to operate differently is to slow down. As music teachers will tell you, the slower you practice the faster you progress. So from that moment I checked the entire references list of every interesting paper. When I found an interesting book in the library, I would read it from beginning to end. When I felt tired I just went to sleep. The world will wait. And guess what? The research world is way more patient than we think (especially in Canada). From that point things started to progress and we completed that paper. As original as I tried to be, several months after our paper a similar paper won the ACCV best paper award. I continued to operate in this more systematic way and it seemed to work for me. In fact, if I had to start over again, I should have read the entire PAMI and IJCV before even touching the keyboard. It would have taken a year, but the alternative is to compete with professors who read every issue of these journals in the last 20 years. After my last paper, a paper draft was brought to my attention that had some ideas in common. That paper was rejected from publication in 2006 because a reviewer couldn't appreciate it.
In this blog, which started as my English practice and turned into a hobby, I tried to describe what graduate students are really going through. It is a documentation of what happened at a special university, in a special city, in a special period in time, seen through the eyes of a person with a peculiar writing style whose opinions I don't always agree with. Some letters were strong, some disappointing, and some drafts (especially on Israelis, lawyers, and women) never posted. One of my personal favorites is actually the first letter. Because when I came to Toronto, I had no clue what I am getting into. For example, I didn't know what living in the same house with people from all continents is like, how cold is -20, or how interesting matrix calculations can be. That letter will always remind me that I learned here something.
