Optimizing Deep Convolutional Neural Network for Facial Expression Recognition

Dinesh Kulkarni,Umesh B Chavan

doi:10.24018/ejers.2020.5.2.495

Abstract

Facial expression recognition (FER) systems have attracted much research interest in the area of Machine Learning. We designed a large, deep convolutional neural network to classify 40,000 images in the data-set into one of seven categories (disgust, fear, happy, angry, sad, neutral, surprise). In this project, we have designed deep learning Convolution Neural Network (CNN) for facial expression recognition and developed model in Theano and Caffe for training process. The proposed architecture achieves 61% accuracy. This work presents results of accelerated implementation of the CNN with graphic processing units (GPUs). Optimizing Deep CNN is to reduce training time for system.

Highlights

F ACIAL expression recognition have found applications in technical fields such as Human-computer-Interaction (HCI) which detect people’s emotions using their facial expressions and security monitoring
The main focus in this work to improve over the existing approach is to use deep learning algorithms to automatic extractions of features
All benchmark in this paper were performed in machine having computation platform with (1) CPU: AMD Phenom II X4 B97 - processor; (2) GPU is GeForce GTX520, compute capability 2.1,48 cores

Summary

Introduction

F ACIAL expression recognition have found applications in technical fields such as Human-computer-Interaction (HCI) which detect people’s emotions using their facial expressions and security monitoring. Use of Machine learning is powerful approach to detect and classify images[1]. To improve their performance, it is necessary to collect larger data-sets, as well as need to build powerful models. The weakest point of machine learning is that it can’t do feature engineering. The biggest drawback is it is time consuming for learning with large data sets with powerful model. GPU deep learning is new model where deep neural networks are trained to recognize patterns from massive amount of data.

Methods

Results

Conclusion