Open AccessOptimized Faster-RCNN in Real-time Facial Expression Classification
Author(s) -
Jiadi Bao,
Shusong Wei,
LV Jing-fan,
Wenli Zhang
Publication year - 2020
Publication title -
iop conference series. materials science and engineering
Language(s) - English
Resource type - Journals
eISSN - 1757-899X
pISSN - 1757-8981
DOI - 10.1088/1757-899x/790/1/012148
Subject(s) - computer science , convolutional neural network , facial expression recognition , logarithm , facial expression , artificial intelligence , raspberry pi , pattern recognition (psychology) , mobile device , gaussian , speech recognition , facial recognition system , embedded system , mathematics , mathematical analysis , physics , operating system , quantum mechanics , internet of things
In order to make convolutional neural network adapt to the mobile terminals which lack of hardware resources in facial expressions recognition. We modified the recent algorithm of real-time CNNs on facial expression. Firstly, we used Gaussian distribution to reduce the irrelevant data. Secondly, we used random forest to reduce the time complexity. We implemented the model and algorithm on raspberry pi. As a result, we reduce the amout of data by about 40% and the time complexity to logarithmic level. Thus, our system can run smoothly on mobile terminals with lack of hardware resources. We validate the accuracy of our system on raspberry pi which has the ability to detect faces and classify the emotion. The accuracy in facial expressions recognition remain stable as the original algorithm. In all, compared with the traditional algorithm, our optimize algorithm improve the number of frames remarkably without reducing the accuracy.