A boosting approach for the simultaneous detection and segmentation of generic objects

Numerous approaches to object detection and segmentation have been proposed in recent years. However, in some general situations these methods are prone to fail due to the nature of the object. For instance, classical approaches to object detection and segmentation obtain good results for some specific object classes (i.e., pedestrian detection or sky segmentation). However, these methods have troubles detecting or segmenting object classes with different distinctive characteristics (i.e., cars and horses versus sky and road). In this paper, we propose a general framework to simultaneously perform object detection and segmentation on objects of different nature. Our approach is based on a boosting procedure which automatically decides - according to the object properties - whether it is better to give more weight to the detection or segmentation process to improve both results. For instance, for some objects, the detection may help to better segment, and viceversa. We validate our approach using different object classes from the well-known LabelMe, TUD and Weizmann databases to obtain competitive detection and segmentation results. Furthermore, our experiments show that the proposed approach is able to correctly annotate new images returned by Internet search engines even when the system is trained with few image examples.