CVPR Daily - Wednesday

DAILY Wednesday Pascal Fua 11 are a good start. We have a centralized admission process, so we get recommendation letters. That helps. Then we interview everyone. Also, we have what we call a fellowship program where some people who look good on paper will be selected. They are funded centrally, not by the lab directly. Over the course of a year, you can tell if they’ve got what it takes. You are obviously still very passionate and excited by what you are doing. How do you transfer this passion and excitement to your students? That’s a very good question. I’d say it sort of goes the other way. If you think back to that tsunami I mentioned, it’s my students I have to thank for not drowning! To be honest, I am fairly resistant to change and it’s hard to be told after 20 years, “You did it all wrong! ” but day after day my students showed me new stuff and eventually, I had to accept that they were probably right. I’m sure you get your passion from them at least as much as you give them their passion. Well, I hope so, but you would have to ask them! That is certainly something that I will do: I bet many of them will be featured in our magazine one day or another. With 300 papers behind you, and many more to come I’m sure, what is the secret to your success? Curiosity is a good start. I have a number of collaborations with people who are not in our field. I work a lot with neuroscientists. When I mentioned the synapses it’s because we have a collaboration with scientists. Now we’ve completely by chance got into the business of computational fluid dynamics, so we have a spin-off that does that. That’s different. If I had done the same thing over and over again for 30 years I would be bored by now. But I’ve been able to change and do different applications, so I’m never bored. Can you share a final thought with our community? I’ll finish with a pet peeve of mine. Our field is driven by benchmarks. Benchmarks are good because you cannot lie, or at least you shouldn’t lie, but at the same time, they are a bit of a straitjacket. That’s particularly true at conferences such as CVPR, where if you want to try something new for which there is no benchmark, the review is going to be that you didn’t compare against whatever is on the benchmark. That is a problem and something that we should think about as a community. There has to be a balance and I’m not sure the current balance is right.