Abstract:
To address high complexity of manually extracted features in traditional methods and insufficient classification of single convolutional neural network(CNN), this paper d...Show MoreMetadata
Abstract:
To address high complexity of manually extracted features in traditional methods and insufficient classification of single convolutional neural network(CNN), this paper designs and implements an ensemble convolutional neural network-based model to improve the recognition rate. The ensemble network model is constructed by training two different subnetwork models and concatenating together the different features extracted from them. Experiments are conducted on the public dataset FER2013, and from the experimental results it is clear that the ensemble network enhances the recognition rate by about 0.4% over the individual network. It is shown that the convolutional neural network model based on the ensemble has a stronger recognition ability.
Date of Conference: 13-14 November 2020
Date Added to IEEE Xplore: 15 July 2021
ISBN Information:
Print on Demand(PoD) ISSN: 2375-141X