Generalization: This paper define a globe energey function for the model, which combines results from sliding window detectors, and low-level pixel-based unary and pairwis relations.
motivation: try to recognition objects, find their location and spatial extent, and also provide the number of instances of objects. This work can be viewed as an integration of object class segmentation methods,which fail to distinguish between adjacent instances of objects of the same class, and object detection approaches, which do not provide information about background classes.