Generalizing over aspect and location for rooftop detection

We present the results of an empirical study in which we evaluated cost-sensitive learning algorithms on a rooftop detection task, which is one level of processing in a building detection system. Speciically, we investigated how well machine learning methods generalized to unseen images that diiered in location and in aspect. For the purpose of comparison, we included in our evaluation a handcrafted linear classiier, which is the selection heuristic currently used in the building detection system. ROC analysis showed that, when generalizing to unseen images that diiered in location and aspect, a naive Bayesian classiier outperformed nearest neighbor and the handcrafted solution.