Related Experiment Video
Updated: Nov 3, 2025
07:07
Author Spotlight: A High-Resolution, Single-Grain, In Vivo Pollen Hydration Bioassay for Arabidopsis thaliana
Published on: June 30, 2023
2.9K
Deep Learning Methods for Improving Pollen Monitoring
Elżbieta Kubera1, Agnieszka Kubik-Komar1, Krystyna Piotrowska-Weryszko2
1Department of Applied Mathematics and Computer Science, University of Life Sciences in Lublin, ul. Głęboka 28, 20-950 Lublin, Poland.
Sensors (Basel, Switzerland)
|June 2, 2021
Summary
Automated pollen identification using deep learning improves allergy risk prediction. A convolutional neural network model achieved 97.88% accuracy in classifying birch, alder, and hazel pollen from microscopic images.
Area of Science:
- Environmental Science
- Botany
- Computer Science
Background:
- Pollen monitoring is crucial for predicting allergy risks.
- Current methods involve manual analysis of Hirst-type sampler slides by specialists.
- Automating pollen identification can enhance monitoring efficiency and accuracy.
Purpose of the Study:
- To develop and evaluate a deep learning method for automated pollen taxon recognition from microscopic images.
- To assess the performance of a custom-built convolutional neural network (CNN) and pre-trained models for this task.
Main Methods:
- A deep CNN model was developed from scratch.
- Publicly available pre-trained deep neural network models were also utilized.
- Models were trained and tested on microscopic images of pollen grains.
Main Results:
- Both custom and pre-trained deep learning models demonstrated effective pollen classification.
- A simple deep learning model achieved high accuracy directly from images.
- The best model reached 97.88% accuracy in distinguishing between birch, alder, and hazel pollen.
Conclusions:
- Deep learning offers a viable solution for automating pollen identification.
- The developed models can support pollen monitoring experts and improve allergy risk assessment.
- Automated systems can significantly enhance the efficiency and precision of aerobiological monitoring.
