Feature selection for object detection: The best group vs. the group of best

The problem of visual object detection, the goal of which is to predict the locations and sizes of all objects of a given visual category (e.g., cars) in a given set of images, is often based on a possibly large set of local features, only a few of which might actually be useful for the given detection setup. Feature selection is concerned with finding a `useful´ subset of features. In this paper, we compare two approaches to feature selection in a visual object detection setup. One of them selects features based on their individual utility scores alone, regardless of possible interdependence with other features. The other approach employs the AdaBoost framework and hence implicitly deals with interdependence. Using two feature extraction methods and several image datasets, we experimentally confirm the significance of feature interdependence: features that perform well individually do not necessarily perform well as a group.