Improving Object Detection Performance Using Scene Contextual Constraints

Contextual information, such as the co-occurrence of objects and spatial relative size among objects, provides rich complex information about digital scenes. It also plays an important role in improving object detection determining out-of-context objects. In this work, we present contextual models that leverage (16 relationships are applied article) to enhance performance two state-of-the-art detectors (i.e., faster RCNN you look only once), which a postprocessing process for most existing detectors, especially refining confidences associated categorical labels, without bounding boxes. We experimentally demonstrate our lead enhancement using common data set used field (MSCOCO), where some experiments, PASCAL2012 is used. show iterating applying enhances further.