This paper presents a novel method for detecting and localizing objects in cluttered real-world scenes, combining object categorization and figure-ground segmentation. The approach uses a flexible learned representation for object shape, combining information from different training examples through a probabilistic extension of the Generalized Hough Transform. This representation allows the system to detect categorical objects in novel images and infer a probabilistic segmentation from the recognition result. The segmentation is then used to improve recognition by focusing on object pixels and discarding background influences. The method also employs a Minimum Description Length (MDL) principle to resolve ambiguities between overlapping hypotheses and account for partial occlusion. Extensive evaluations on large datasets show that the system is effective for various object categories, including rigid and articulated objects, and achieves competitive performance with smaller training sets compared to comparable systems.This paper presents a novel method for detecting and localizing objects in cluttered real-world scenes, combining object categorization and figure-ground segmentation. The approach uses a flexible learned representation for object shape, combining information from different training examples through a probabilistic extension of the Generalized Hough Transform. This representation allows the system to detect categorical objects in novel images and infer a probabilistic segmentation from the recognition result. The segmentation is then used to improve recognition by focusing on object pixels and discarding background influences. The method also employs a Minimum Description Length (MDL) principle to resolve ambiguities between overlapping hypotheses and account for partial occlusion. Extensive evaluations on large datasets show that the system is effective for various object categories, including rigid and articulated objects, and achieves competitive performance with smaller training sets compared to comparable systems.