Tuesday, January 01, 2008
vision and machine learning
I was thinking about the dominant practices in computer vision. It is very unpleasant to acknowledge some of the usual trends. The worst of these are the ones related to adoption/application of existing machine learning techniques. There are, undoubtedly, many contributions in machine learning from "vision researchers", and there are numerous useful applications of learning algorithms to solve vision related problems. The point of concern is the growing negligence of the "old school" vision approaches such as physics and geometry based vision. The idea of solving real problems is becoming obsolete; building models is the new fashion statement. Is this the standard dilemma of systems versus algorithms? I think both of these perspectives are good ideally, but the implementations are not perfect: we end up with non-generalizable systems (even worse, with non-public implementations) and incremental algorithms. How are these useful for the community? Other than recognition (in the form of publications), what purpose are these papers serving? This raises a question: what is the distribution of the number of citations of papers from conferences like CVPR, ICCV and ECCV? A useful script to work on.
Subscribe to:
Post Comments (Atom)
Posts
2 comments:
I don't think the two are mutually exclusive, necessarily.
A good probabilistic model should incorporate everything you know about the problem, and if that includes physical/geometric details, so be it. There's still benefit to formulating the problem probabilistically, as you can then account for uncertainty in a principled way; sort through several plausible interpretations and do your best from both angles - old school vision and machine learning/statistics - to decide which one is closest to the truth.
David, I agree with you that the two fields are related and that we should use probabilistic modeling and other available abstractions. Raising a doubt about the utility of one over the another is not my concern. On the contrary, my complaint is about the recent craze of blindly applying machine learning models/approaches to vision problems without any reasonable consideration for the vision aspects. I believe if we adapt the machine learning tools in a fashion appropriate to the problem at hand and incorporating the "old school" vision knowledge into these models, we get more insight into the problem and a possible solution, as opposed to considering old school vision and machine learning as two competing approaches and in isolation.
Post a Comment