Related Experiment Video
Updated: Jul 10, 2025
09:10
Multimodal Analysis of Microplastics in Drinking Water using a Silicon Nanomembrane Analysis Pipeline
Published on: June 13, 2025
91
Efficient and accurate microplastics identification and segmentation in urban waters using convolutional neural
1School of Civil Engineering and Transportation, State Key Laboratory of Subtropical Building and Urban Science, South China University of Technology, Guangzhou 510641, China; Pazhou Lab, Guangzhou 510335, China.
The Science of the Total Environment
|November 24, 2023
Summary
Convolutional neural networks (CNNs) efficiently identify microplastics (MPs) in urban waters. UNet and UNet2plus models offer accurate and rapid detection, crucial for large-scale pollution studies.
Area of Science:
- Environmental Science
- Computer Science
- Data Science
Background:
- Microplastics (MPs) are pervasive pollutants in urban waters, posing risks to ecosystems and human health.
- MPs act as vectors for other pollutants, exacerbating global contamination.
- Manual identification of MPs is labor-intensive and time-consuming, hindering large-scale research.
Purpose of the Study:
- To develop and evaluate convolutional neural network (CNN) models for efficient and accurate microplastic identification in urban waters.
- To compare the performance of UNet, UNet2plus, and UNet3plus models in segmenting microplastics from complex backgrounds.
Main Methods:
- A dataset of microplastics from urban waters in southern China was curated.
- UNet, UNet2plus, and UNet3plus CNN models were trained and validated on the microplastic dataset.
- Computational and inference performance, including segmentation accuracy (mIoU), was evaluated for each model.
Main Results:
- UNet and UNet2plus achieved high segmentation accuracy (mIoU ≈ 91.45% and 91.08%, respectively).
- All three models provided efficient inferences, with UNet and UNet2plus completing 100 images in under 1 and 2 seconds.
- UNet3plus showed lower performance (mIoU ≈ 82.21%) compared to UNet and UNet2plus.
Conclusions:
- UNet and UNet2plus models demonstrate significant potential for accurate and efficient microplastic identification in urban aquatic environments.
- These CNN models can substantially reduce manual effort in large-scale microplastic pollution monitoring.
- Further development of AI-driven solutions is crucial for effective management of microplastic contamination.
Keywords:
Convolutional neural networkEfficiency and accuracyIdentification and segmentationMicroplasticsUrban waters
