Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Video

Updated: Aug 29, 2025

Additive Manufacturing-Enabled Low-Cost Particle Detector
06:05

Additive Manufacturing-Enabled Low-Cost Particle Detector

Published on: March 24, 2023

1.3K

Automated particle recognition for engine soot nanoparticles.

E Haffner-Staton1, L Avanzini1, A La Rocca1

  • 1Department of Mechanical, Materials and Manufacturing Engineering, University of Nottingham, University Park, Nottinghamshire, UK.

Journal of Microscopy
|September 6, 2022
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Evidence-based multidisciplinary model of care for newborn screening in spinal muscular atrophy.

Genetics in medicine : official journal of the American College of Medical Genetics·2026
Same author

The effect of a digital training tool to aid chest image interpretation: Hybridising eye tracking technology and a decision support tool.

Radiography (London, England : 1995)·2020
Same author

Digital training platform for interpreting radiographic images of the chest.

Radiography (London, England : 1995)·2018
Same author

A matched comparison study of hepatitis C treatment outcomes in the prison and community setting, and an analysis of the impact of prison release or transfer during therapy.

Journal of viral hepatitis·2016
Same author

BRCA1, BRCA2 and PALB2 mutations and CHEK2 c.1100delC in different South African ethnic groups diagnosed with premenopausal and/or triple negative breast cancer.

BMC cancer·2015
Same author

Establishing comprehensive oral assessments for children with safeguarding concerns.

British dental journal·2015
Same journal

In operando imaging of the space-charge region in a 4H-SiC MOSCAP using STEM-EBIC.

Journal of microscopy·2026
Same journal

The future of DXA: How AI is transforming bone health diagnostics.

Journal of microscopy·2026
Same journal

The Origins of Ploem's Filter Cube: A Pandora's Box.

Journal of microscopy·2026
Same journal

The reproducibility gap in graph neural network workflows for cell dynamics: A checklist-driven case study.

Journal of microscopy·2026
Same journal

Assessing the reproducibility of a bioimage analysis workflow characterising tissue flow in Drosophila.

Journal of microscopy·2026
Same journal

Modular training resources for bioimage analysis.

Journal of microscopy·2026
See all related articles

Convolution neural networks (CNNs) efficiently classify soot and non-soot nanoparticles in TEM images. ResNet18 offers high accuracy with significantly reduced training time compared to ResNet50.

Area of Science:

  • Materials Science
  • Nanotechnology
  • Computer Science

Background:

  • Accurate classification of carbon nanoparticles is crucial for understanding combustion processes and material properties.
  • Manual classification of transmission electron microscopy (TEM) images is time-consuming and subjective.
  • Deep learning offers a potential solution for automated and objective image analysis.

Purpose of the Study:

  • To evaluate the effectiveness of a pre-trained Convolutional Neural Network (CNN), specifically Residual Network (ResNet), for classifying soot and non-soot carbon nanoparticles in TEM images.
  • To compare the performance and efficiency of different ResNet architectures (ResNet18 and ResNet50) and training dataset sizes.
  • To determine the optimal training parameters for accurate and rapid classification.

Main Methods:

Keywords:
TEMautomotivenanoparticlesneural networkssootvision learning

More Related Videos

Measuring Sub-23 Nanometer Real Driving Particle Number Emissions Using the Portable DownToTen Sampling System
08:59

Measuring Sub-23 Nanometer Real Driving Particle Number Emissions Using the Portable DownToTen Sampling System

Published on: May 22, 2020

5.6K
Multimodal Analysis of Microplastics in Drinking Water using a Silicon Nanomembrane Analysis Pipeline
09:10

Multimodal Analysis of Microplastics in Drinking Water using a Silicon Nanomembrane Analysis Pipeline

Published on: June 13, 2025

432

Related Experiment Videos

Last Updated: Aug 29, 2025

Additive Manufacturing-Enabled Low-Cost Particle Detector
06:05

Additive Manufacturing-Enabled Low-Cost Particle Detector

Published on: March 24, 2023

1.3K
Measuring Sub-23 Nanometer Real Driving Particle Number Emissions Using the Portable DownToTen Sampling System
08:59

Measuring Sub-23 Nanometer Real Driving Particle Number Emissions Using the Portable DownToTen Sampling System

Published on: May 22, 2020

5.6K
Multimodal Analysis of Microplastics in Drinking Water using a Silicon Nanomembrane Analysis Pipeline
09:10

Multimodal Analysis of Microplastics in Drinking Water using a Silicon Nanomembrane Analysis Pipeline

Published on: June 13, 2025

432
  • Applied pre-trained ResNet (18 and 50 layers) to classify soot and non-soot nanoparticles in TEM images.
  • Trained networks using varying dataset sizes (100, 400, 1400 images) and assessed performance on an independent test set (200 images).
  • Conducted fivefold cross-validation experiments on a larger dataset (1600 images) to validate classification accuracy and efficiency.

Main Results:

  • ResNet18 and ResNet50 achieved statistically similar classification accuracies, with ResNet18 requiring 25-35% less training time.
  • Classification accuracy increased with training set size, reaching 95% with 1400 images.
  • Cross-validation demonstrated classification accuracy exceeding 91% for both ResNet architectures.
  • The most efficient method, ResNet18 trained for 5 epochs, achieved 91.2% accuracy in 84 seconds for 1600 images.

Conclusions:

  • CNNs, particularly ResNet, are highly effective for automated classification of soot and non-soot nanoparticles in TEM images.
  • ResNet18 provides a computationally efficient alternative to ResNet50 with comparable accuracy.
  • Automated classification using CNNs significantly reduces analysis time compared to manual methods, especially for large datasets.