In this paper, we propose a convolution neural network structure optimized for facial recognition.
We optimized the structure of the hidden layer to reduce the amount of learning computation, and as a result, the learning speed was improved. In ad...
In this paper, we propose a convolution neural network structure optimized for facial recognition.
We optimized the structure of the hidden layer to reduce the amount of learning computation, and as a result, the learning speed was improved. In addition, we improved the emotional recognition accuracy through learning parameters and data argumentation. As a result of the final learning experiment, the predicated rate was 84.9% and the learning speed was improved compared to VGG network [1].