Object Detection based on Deformable Parts Model with Global Context

The Deformable part-based model (DPM) is a remarkable algorithm in object detection. In this paper, it is combined with the global information to improve its performance. The gist feature of an image is extracted to capture its global information. After that, the principal component analysis (PCA) is used to reduce the dimensionality of the gist feature. The k nearest neighbor distance (k-NND) is utilized to judge the similarity of an image and an object. To accelerateour algorithm, every object is represented as several object models, which can be obtained by affinity propagation (AP). Finally, the score of DPM is merged with one of k-NND to rescore an image. Experimental results show that the introduction of global context is positive for the object detection.