Inter-active learning of randomized tree ensembles for object detection

The detection of multiple objects in noisy images without an explicit model is one of the most challenging tasks in computer vision. In this paper we propose a novel object detection algorithm, termed inter-active tree ensemble (ITE), which can be applied in an off-the-shelf manner to a large variety of tasks. The contributions of this work are twofold: (i) we describe a feature basis which is able to capture shape information and which is illumination invariant. Furthermore, the feature set is guaranteed to be generally applicable due to its high flexibility, (ii) we present an interactive ensemble learning algorithm based on randomized trees, which can cope with exceptionally high dimensional feature spaces in an efficient manner. Contrary to classical approaches, subspaces are not split based on thresholds but by learning relations between features. ITE compares favorably to state of the art methods and we demonstrate its performance on a real world problem in computational pathology.