Related Experiment Video
Updated: Jul 13, 2026
06:49
Evaluation of a Smartphone-based Human Activity Recognition System in a Daily Living Environment
Published on: December 11, 2015
8.9K
An AI based smart-phone system for asbestos identification.
Michael Rolfe1, Samantha Hayes2, Meaghan Smith1
1Department of Chemistry and Biotechnology and Department of Computing Technologies, School of Science, Computing and Engineering Technologies, Swinburne University of Technology Melbourne, VIC 3122, Australia.
Journal of Hazardous Materials
|November 2, 2023
Summary
This study developed a smartphone-based image recognition system for asbestos identification. The system achieved 90% accuracy, offering a portable and cost-effective solution for detecting asbestos materials.
Area of Science:
- Environmental Science
- Materials Science
- Computer Science
Background:
- Asbestos identification is crucial for environmental and economic reasons.
- Current methods rely on laboratory analysis with light microscopy and specialized mounting.
- A need exists for more accessible and portable asbestos detection methods.
Purpose of the Study:
- To develop a smartphone-based image recognition system for asbestos identification.
- To evaluate the effectiveness of portable microscopy combined with deep learning for this task.
- To compare the performance of different convolutional neural network (CNN) models.
Main Methods:
- Utilized a portable 30x microscope with a smartphone camera.
- Trained a deep learning model using 7328 images from over 1000 asbestos cement sheet samples.
- Tested and compared three CNN models: ResNet101, InceptionV3, and VGG_16.
Main Results:
- ResNet101 achieved the highest accuracy (98.46%) with a loss of 3.8%.
- The phone-based system correctly identified asbestos distinctiveness 90% of the time without specialized mounting.
- ResNet101 demonstrated superior performance compared to other tested deep learning models.
Conclusions:
- A portable smartphone-based system can effectively identify asbestos.
- Deep learning, particularly ResNet101, offers a promising approach for accurate asbestos detection.
- This technology provides a more accessible and potentially cost-effective alternative to traditional laboratory methods.
