Compensating for visually missing features: Scale adaptive recognition of objects using probabilistic voting

In this work-in-progress paper, we present an efficient methodology for a scale-adaptive recognition of objects. We introduce a new object recognition approach, which detects an object in a scene while probabilistically predicting visually missing features. The idea is to enable a better recognition by considering the fact that object features may not be detected depending on its situation (e.g. distance and occlusion). A probabilistic voting-based methodology is developed.