TCAS: A multiclass object detector for robot and computer vision applications Rodrigo Verschae, Javier Ruiz-del-Solar Advanced Mining Technology Center, Universidad de Chile, Chile Abstract. Building efficient object detection systems is an important goal of computer and robot vision. If several object types are to be detected, the most simple solution is to run several object-specific classifiers independently of each other (in parallel). This solution is computationally expensive if several object classes are to be detected. In this paper, TCAS, a new classifier structure designed to be used on multiclass object detection problems is introduced as an alternative solution. TCAS offers an efficient solution and reduces the aggregated false de- tection rate. TCAS extends cascade classifiers (introduced by Viola & Jones) to the multiclass case and corresponds to a nested coarse-to-fine tree of multiclass nested boosted cascades. Results for three different object detection problems are presented: face and hand detection, robot detection, and multiview face de- tection. In the experiments, the obtained TCAS have classification times about 2-times shorter than the ones obtained using parallel cascades, and have the same or lower number of false positives (for the same detection rate). 1 Introduction The development of robust and real-time object detection systems is an important goal of the robot-vision and computer-vision communities. Applications of multiclass de- tection systems include human-robot interaction (e.g. face and hand detection), and autonomous drivings system (e.g. car and sign detection), among many others. For detecting all instances of an object-class, the sliding window approach [1] re- quires to perform an exhaustive search over image patches at different scales and posi- tions. The classification of all possible windows (patches) may require a high computa- tional power. To achieve an efficient detection, less processing time should be spent on non-object windows than on object windows. This can be achieved using cascade clas- sifiers [1] (Fig. 1 (a)). Starting from the seminal work of Viola & Jones [1] on cascade boosted classifiers for object detection, several improvements have been proposed, with the introduction of nested cascades [2] being one of the most important ones. In order to detected multiple classes of objects, the most simple solution is to use several cascades in a a concurrent manner, with each cascade trained for each particular class independently of each other. The main problem with this approach is that the pro- cessing time and the false positive rate (fpr) aggreegates with the number of classes. As a solution to these problems we propose a multiclass classifier with a tree structure. The proposed detector corresponds to a multiclass nested tree of multiclass nested cascades that compared to the use of parallel cascades presents a considerable gain in processing time. Results in 3 different detection problems are presented later.