Learning-based methods in computer vision make use of training data to build systems for visual analysis. For example, one may train a system for detecting faces using training images of faces. Training data is often given in the forms of image or video collections, together with target labels. Such data is often fed into a machine learning algorithm, that will learn to predict such labels given novel images or video. Learning-based methods have been used for a variety of computer vision tasks, including low-level problems such as image-denoising, and high-level tasks such as object recognition and scene classification.