Saturday, December 27, 2008
When pre-computing a similarity matrix, there there is a large cost in both time and memory for computing pairwise similarity between every sample, O(N^2). There are two main ways to slightly reduce the cost of either one of these demands... Approximate Hashing: Hashing projects a set of features into a lower dimensional representation while simultaneously […]
Tired of people mocking your profile photos? Don't have photographic evidence that you were a child? If so, just crop out a bunch of your faces and then mathematically combine them for others' viewing [dis]pleasure. Start with any photos you want, cut out your face (roughly) and paste to a new file. Tiff is the […]
Congratulations, you found the only digital home page for Eric Zavesky, a principal member of technical staff at AT&T Labs Research. I research retrieval and interaction problems in a group focusing on Video and Multimedia Retrieval Services. Specifically, I'm interested in shifting the cognitive burden of modern indexing and retrieval systems (i.e. how an average […]