A dynamically trained neural network is proposed in this paper proper for adapting the network performance to non stationary image or video inputs. The sheme includes, on one hand, a retrieval mechanism which selects the most approprite network from the system memory and, on the other hand, a weight perturbation procedure which adapts the network weights to the current condition. If no suitable network exists in memory, new weights and network structure are crated and them stored for future use. Experimental results are provided indicating the good performance of the proposed system to computer and machine vision applications.
JavaScript jest wyłączony w Twojej przeglądarce internetowej. Włącz go, a następnie odśwież stronę, aby móc w pełni z niej korzystać.