News for hackers. Stuff that matters.
Re:Large Scale Visual Recognition Challenge 2011 - Results by pmelendez on Thursday October 04, 2012 @01:28PM (#4613349)
"So you're asserting that the 10% improvement by Supervision is because they used the raw RGB pixels. Is that right?"

No, what I meant is that SuperVision did very well because their feature space is richer than the other teams, but IMHO that is resulting for the deep convolutional process which it used to generate rich features. This is a good explanation of the subject:

http://ai.stanford.edu/~ang/papers/icml12-HighLevelFeaturesU...

I deleted all the comments I could as they were complaining about the title. Since it was changed to a proper one most of my comments are not relevant any more.

There are no comments on this comment yet.