A scene-specific deformable part-based model for object detection

The various scale makes detecting and localizing objects a challenging problem, especially for small-scale instances [1, 2]. While most existing models focus on detection in static images, we investigate the static video surveillance scenario. In this paper, a probabilistic graphical model is proposed to integrate a local generic object detector and scene-specific contextual features. The proposed model outperforms most part-based models by extending them into a multiresolution structure. Experimental results on the public dataset CAVIAR [3] demonstrate that our model surpasses the conventional deformable part-based model (DPM) with an improvement of 28.25% in the average precision. In addition, our model can be easily adapted to a new scenario without a re-training process.