ニューラルネットワークによる若年者顔画像からの年齢層推定 : モザイク特徴とKL特徴の比較

概要

論文の詳細を見る
This paper describes the estimation of the age group based on the facial images of young men by using the neural network. First, we analyzed the data which measured the facial shape. The facial shape had been gradually changing from the infant age, and the age that the shape was completed was about 22 years old in the male, and about 15 years old in the female. Then, we experimented by using the facial images of male with a long changing span. Their ages were from 12 to 22 years old. The input data to the neural network used the mosaic features and the KL features. The neural network was able to classify those facial images into four age groups by using those features. Those identification rates were about 80%. The result of classifier which had used the KL feature was similar to the subjective impression which the person had evaluated. Moreover, the facial feature of each age group has been extracted by analyzing the connection weights between the hidden-layer unit and the input-layer units. The feature of each age group which had been obtained by the mosaic feature was a facial outline. The eigenface extracted by the KL feature showed the facial feature of each age group.
社団法人映像情報メディア学会の論文
2000-03-23