Ensemble of Deep Neural Networks for Estimating Particulate Matter from Images

Particulate matter with diameters less than 2.5 micrometers (PM2.5) is one of the most common air pollutants and may cause many severe diseases. An efficient PM2.5 monitoring system is of great benefit for human health and air pollution control. In this paper, we estimate PM2.5 concentrations using...

Full description

Saved in:
Bibliographic Details
Published in2018 IEEE 3rd International Conference on Image, Vision and Computing (ICIVC) pp. 733 - 738
Main Authors Rijal, Nabin, Gutta, Ravi Teja, Cao, Tingting, Lin, Jerry, Bo, Qirong, Zhang, Jing
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.06.2018
Subjects
Air pollution
Biological neural networks
Convolution
convolutional neural network
deep learning
ensemble
estimation
Machine learning
Neurons
particulate matter
Training
Online AccessGet full text

Cover

More Information
Summary:Particulate matter with diameters less than 2.5 micrometers (PM2.5) is one of the most common air pollutants and may cause many severe diseases. An efficient PM2.5 monitoring system is of great benefit for human health and air pollution control. In this paper, we estimate PM2.5 concentrations using outdoor images by a proposed ensemble of deep neural networks-based regression, which uses a feedforward neural network to combine the PM2.5 predictions yielded by three convolutional neural networks, VGG-16, Inception-v3, and ResNet50, and calculate the final PM2.5 prediction of the image. A PM2.5 image dataset with 1460 photos was used for performance evaluation. The experimental results demonstrated that the proposed ensemble can provide more accurate PM 2.5 estimation than all three individual deep learning networks used and, therefore, can be used for image-based PM 2.5 monitoring.
DOI:10.1109/ICIVC.2018.8492790