Ensemble of deep neural networks with probability-based fusion for facial expression recognition

G Wen, Z Hou, H Li, D Li, L Jiang, E Xun - Cognitive Computation, 2017 - Springer
G Wen, Z Hou, H Li, D Li, L Jiang, E Xun
Cognitive Computation, 2017Springer
Convolutional neural network (CNN) is a very effective method to recognize facial emotions.
However, the preprocessing and selection of parameters of these methods heavily depend
on the human experience and require a large amount of trial-and-errors. This paper
presents an ensemble of convolutional neural networks method with probability-based
fusion for facial expression recognition, where the architecture of CNN was adapted by
using the convolutional rectified linear layer as the first layer and multiple hidden maxout …
Abstract
Convolutional neural network (CNN) is a very effective method to recognize facial emotions. However, the preprocessing and selection of parameters of these methods heavily depend on the human experience and require a large amount of trial-and-errors. This paper presents an ensemble of convolutional neural networks method with probability-based fusion for facial expression recognition, where the architecture of CNN was adapted by using the convolutional rectified linear layer as the first layer and multiple hidden maxout layers. It was constructed by randomly varying parameters and architecture around the optimal values for CNN, where each CNN as the base classifier was trained to output a probability for each class. These probabilities were then fused through the probability-based fusion method. The conducted experiments on benchmark data sets validated our method, which had better accuracy than the compared methods. The proposed method was novel and efficient for facial expression recognition.
Springer
以上显示的是最相近的搜索结果。 查看全部搜索结果