Rotation, scaling, shift, and distortion-invariant target recognition

This model introduces rotation invariance into the Neocognitron model by way of rotation layers. The most likely orientation of the target is detected and used for further pattern classification using an algorithm similar to that for the Neocognitron model. The visual target recognition capability of this model is demonstrated.