Translational invariant object recognition using back propagation network

Summary form only given, as follows. A novel method using neural networks for invariant object recognition has been developed. The objective is to permit the recognition of objects in any shifted position while the objects are presented to the network in only one standard location during the training procedure. The presence of multiple objects and noise corruption in the scene is permitted. This method utilizes the secondary responses activated by the hypernetwork, and a confirmative network is used to obtain the object identification and location, based on these secondary responses