Related Experiment Video
Updated: Sep 20, 2025
07:26
In Situ Microscopy for Real-time Determination of Single-cell Morphology in Bioprocesses
Published on: December 5, 2019
8.0K
Nanoparticle Detection on SEM Images Using a Neural Network and Semi-Synthetic Training Data
Jorge David López Gutiérrez1, Itzel Maria Abundez Barrera1, Nayely Torres Gómez1
1División de Estudios de Posgrado e Investigación, Instituto Tecnológico de Toluca, Tecnológico Nacional de México, Av. Tecnológico s/n, Colonia Agrícola Bellavista, Metepec, México City 52149, Mexico.
Nanomaterials (Basel, Switzerland)
|June 10, 2022
Summary
This study introduces YOLO neural networks for detecting nanoparticles in scanning electron microscopy images. The models show promise in identifying particles, even in challenging, close-proximity scenarios.
Area of Science:
- Materials Science
- Computer Science
- Nanotechnology
Background:
- Image processing is crucial for analyzing nanoparticle data from electron microscopy.
- Deep learning, specifically neural networks, has emerged as effective for image analysis in microscopy.
- Existing methods often require manual intervention for nanoparticle detection and measurement.
Purpose of the Study:
- To develop and evaluate YOLO neural network models for detecting cubical and quasi-spherical nanoparticles in scanning electron microscopy (SEM) images.
- To assess the models' performance on diverse datasets, including those with synthetic images.
- To identify challenges in nanoparticle detection, such as close proximity.
Main Methods:
- Four detection models were developed using two versions of the YOLO (You Only Look Once) neural network architecture.
- Training datasets comprised a mix of real SEM images and synthetically generated nanoparticle images.
- Models were trained to detect specific nanoparticle shapes (cubical and quasi-spherical).
Main Results:
- The trained YOLO models demonstrated the capability to detect nanoparticles in SEM images not previously encountered during training.
- The models showed success in identifying nanoparticles, though challenges persisted in distinguishing closely spaced particles.
- Performance varied depending on the nanoparticle shape and image complexity.
Conclusions:
- YOLO neural networks offer a viable automated approach for nanoparticle detection in SEM imagery.
- The integration of synthetic data aids in training robust detection models.
- Further refinement is needed to overcome limitations posed by nanoparticle aggregation and close proximity.
