From a list of undergraduate projects available at the Center for Intelligent machines at McGill, "Content-Based Image Retrieval":
2. Finding Animals in Internet Imagery Using Learning MethodsThis project, based on a recently published paper, involves finding images on the Internet that contain animals. Recently Viola et al. have introduced a rapid object detection scheme based on a boosted cascade of simple feature classifiers. In a more current paper, two extensions to their approach were introduced: Firstly, a novel set of rotated Haar-like features was proposed. These features significantly enrich the original simple features and can also be calculated efficiently. Secondly, based on a thorough analysis of different boosting algorithms (namely, Discrete, Real and Gentle Adaboost), it was found that the Gentle Adaboost learning method was the superior training approach. The complete object detection training and detection system is available in the Open Computer Vision Library atsourceforge.net. Therefore, the primary objective of this project is to use this software to implement the search for animals.
Far less funny are numbers 1 ("Keeping Track of Humans and Objects in Videos for Surveillance Applications") and 3 ("Detecting Moving Objects on the Street Using a Color Video Surveillance Camera"), which are basically top-down surveillance research. Shame on you Mr. Levine.
In any case, I am a little surprised to realize this stuff is going on here. I'm especially intrigued by the "Shared Reality Environment" Lab. Where do these people hang out and socialize? ;)
