Connectivity Strategies for Second-Order Rotation Invariant Neural Networks

T.-N. Yang, S.-J. Yen, C.-J. Chen, and C.-J. Lee (Taiwan)


: neural networks; invariant recognition; rota tion invariant features


A second-order neural network is designed to be invari ant to changes in rotation. Rotation invariance is achieved through a special arrangement of the network structure. The training set only requires one view of each target ob ject. We describe the weight sharing strategy and present a coin recognition neural network illustrating its usefulness. The simulation results show that the proposed neural net work can distinguish between the two target coins indepen dent of the transformation in rotation.

Important Links:

Go Back