Related Experiment Video
Updated: Aug 18, 2025
05:31
Sampling, Sorting, and Characterizing Microplastics in Aquatic Environments with High Suspended Sediment Loads and Large Floating Debris
Published on: July 28, 2018
16.1K
Deep learning based approach for automated characterization of large marine microplastic particles
Xiao-Le Han1, Ning-Jun Jiang2, Toshiro Hata3
1Institute of Geotechnical Engineering, Southeast University, Nanjing, Jiangsu, China; Department of Civil and Environmental Engineering, University of Hawaii at Manoa, Honolulu, HI, USA.
Marine Environmental Research
|December 10, 2022
Summary
This study introduces Mask R-CNN, a deep learning method for identifying marine microplastics. It accurately locates, classifies, and segments plastic particles, improving upon traditional visual inspection methods.
Area of Science:
- Environmental Science
- Marine Biology
- Computer Science
Background:
- Marine microplastic pollution is a significant global environmental concern.
- Current visual inspection methods for microplastics have high misidentification rates.
- Advanced automated methods are needed for accurate microplastic characterization.
Purpose of the Study:
- To develop and validate a deep learning approach for automated marine microplastic characterization.
- To assess the performance of the Mask R-CNN algorithm in locating, classifying, and segmenting microplastics.
- To compare the proposed method with existing techniques like U-Net.
Main Methods:
- A deep learning model, Mask R-CNN with a Resnet 101 backbone, was employed.
- A dataset of 3000 microplastic images was created for training and validation.
- The algorithm's performance was evaluated against various backgrounds and compared with U-Net.
Main Results:
- The Mask R-CNN algorithm achieved high performance metrics: Precision=93.30%, Recall=95.40%, F1 score=94.34%.
- Excellent performance was also observed in Average Precision for bounding box (APbb=92.7%) and mask (APm=82.6%).
- The algorithm demonstrated a processing speed of 12.5 FPS and could be tuned in under 8 hours.
Conclusions:
- The Mask R-CNN algorithm shows significant promise for accurate and efficient marine microplastic characterization.
- This deep learning approach offers a potential solution for large-scale microplastic surveys.
- The study highlights the effectiveness of AI in addressing environmental pollution challenges.
