Dynamic tree map for video analysis

Abstract

We introduce the DTM (dynamic genetic tree-map), a self-organizing neural network capable of structuring the optimal features in the data destined for recognition. The DTM uses a genetic algorithm to determine the degree of importance of features that have not been considered in the existing neural network. And we apply the GTM (genetic tree-map) that includes tree structure according to the precedence of the features. We suggest that the neurons inside the neural network are dynamically separated and incorporated based on the extended method of the DTM.