Description and recognition of complex spatial configurations of object pairs with Force Banner 2D features

A major challenge in scene understanding is the handling of spatial relations between objects or object parts. Several descriptors dedicated to this task already exist, such as force histogram which a typical example relative position descriptor. By computing interaction two for given all directions, it gives good overview configuration, and has useful properties that can make invariant 2D viewpoint. Considering using complementary forces (negative repulsion, positive attraction) should improve description complex configurations, we propose extend panel so more complete This descriptor called “(discrete) Force Banner” be used input classical Convolutional Neural Network (CNN), benefiting from their powerful performances, reduced into compact features use them another system. As an illustration its ability describe solve classification problem aiming discriminate simple relations, but with variable configuration complexities. Experimental results obtained on datasets synthetic natural images various shapes highlight interest approach, particular configurations.