Deep learning for the automation of particle analysis in catalyst layers for polymer electrolyte fuel cells

Colliard-Granero, André; Batool, Mariah; Janković, Jasna; Jitsev, Jenia; Eikerling, Michael; Malek, Kourosh; Eslamibidgoli, Mohammad Javad

doi:10.1039/d1nr06435e

Cited by 17 publications

(13 citation statements)

References 66 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Although the original designs of such proposal-free instance segmentation algorithms are demonstrated with optical microscopy images of aggregated cells as examples, , several recent studies have employed them to analyze SEM and TEM images. For example, the StarDist has been demonstrated to give an accurate measurement of the particle size distribution of the Pt catalysts supported on carbon for polymer electrolyte fuel cells, by training on datasets as small as 30 images. Such imaging-based characterization of a catalyst enables the high-throughput screening in the design and fabrication of novel materials.…”

Section: Application Of ML In Em Data Analysismentioning

confidence: 99%

Electron Microscopy Studies of Soft Nanomaterials

et al. 2023

View full text Add to dashboard Cite

This review highlights recent efforts on applying electron microscopy (EM) to soft (including biological) nanomaterials. We will show how developments of both the hardware and software of EM have enabled new insights into the formation, assembly, and functioning (e.g., energy conversion and storage, phonon/photon modulation) of these materials by providing shape, size, phase, structural, and chemical information at the nanometer or higher spatial resolution. Specifically, we first discuss standard real-space two-dimensional imaging and analytical techniques which are offered conveniently by microscopes without special holders or advanced beam technology. The discussion is then extended to recent advancements, including visualizing three-dimensional morphology of soft nanomaterials using electron tomography and its variations, identifying local structure and strain by electron diffraction, and recording motions and transformation by in situ EM. On these advancements, we cover state-of-the-art technologies designed for overcoming the technical barriers for EM to characterize soft materials as well as representative application examples. The even more recent integration of machine learning and its impacts on EM are also discussed in detail. With our perspectives of future opportunities offered at the end, we expect this review to inspire and stimulate more efforts in developing and utilizing EM-based characterization methods for soft nanomaterials at the atomic to nanometer length scales in academic research and industrial applications.

show abstract

Section: Application Of ML In Em Data Analysismentioning

confidence: 99%

Electron Microscopy Studies of Soft Nanomaterials

et al. 2023

View full text Add to dashboard Cite

show abstract

“…5). 15,16,17 Standard architectures for instance segmentation involve fully convolutional networks which first perform convolution and down-sampling operations to extract the features (encoder), followed by up-sampling and transpose convolution operations (decoder) until the starting input size is reached. The U-Net model, in addition to the described data flow, concatenates the up-sampled decoder features with the corresponding ones from the encoder (Fig.…”

Section: Upper-limit Floatmentioning

confidence: 99%

UTILE-Gen: Automated Image Analysis in Nanoscience Using Synthetic Dataset Generator and Deep Learning

Colliard-Granero

Jitsev

Eikerling

et al. 2023

Preprint

View full text Add to dashboard Cite

This work presents the development and implementation of a deep learning-based workflow for autonomous image analysis in nanoscience. A versatile, agnostic, and configurable tool was developed to generate instance-segmented imaging datasets of nanoparticles. The synthetic generator tool employs domain randomization to expand the image/mask pairs dataset for training supervised deep learning models. The approach eliminates tedious manual annotation and allows training of high-performance models for microscopy image analysis based on convolutional neural networks. We demonstrate how the expanded training set can significantly improve the performance of the classification and instance segmentation models for a variety of nanoparticle shapes, ranging from spherical-, cubic-, rod-shaped to amorphous nanoparticles. Finally, the trained models were deployed in a cloud-based analytics platform for the autonomous particle analysis of microscopy images.

show abstract

“…The application of DL algorithms for the analysis of EM data has been thoroughly reviewed in a recent article . For example, the U-Net architecture with StarDist formulation for loss function were trained to automate the particle size distribution analysis of electrocatalyst materials, where various shape, texture, and patterns are generated between the overlapping catalyst NPs and the support material . DL models were also used for real-time segmentation of NPs in liquid phase EM movies to statistically examine the diffusion, reactivity, and assembly kinetics of cube-, prism-, and rod-shaped colloidal NPs .…”

Section: Introductionmentioning

confidence: 99%

UTILE-Gen: Automated Image Analysis in Nanoscience Using Synthetic Dataset Generator and Deep Learning

et al. 2023

Self Cite

View full text Add to dashboard Cite

This work presents the development and implementation of a deep learning-based workflow for autonomous image analysis in nanoscience. A versatile, agnostic, and configurable tool was developed to generate instance-segmented imaging datasets of nanoparticles. The synthetic generator tool employs domain randomization to expand the image/mask pairs dataset for training supervised deep learning models. The approach eliminates tedious manual annotation and allows training of high-performance models for microscopy image analysis based on convolutional neural networks. We demonstrate how the expanded training set can significantly improve the performance of the classification and instance segmentation models for a variety of nanoparticle shapes, ranging from spherical-, cubic-, to rod-shaped nanoparticles. Finally, the trained models were deployed in a cloud-based analytics platform for the autonomous particle analysis of microscopy images.

show abstract

Deep learning for the automation of particle analysis in catalyst layers for polymer electrolyte fuel cells

Abstract: This paper presents a deep learning-based approach to automate particle size analysis in the microscopy images of catalyst layers for polymer electrolyte fuel cells.

Cited by 17 publications

References 66 publications

Electron Microscopy Studies of Soft Nanomaterials

Electron Microscopy Studies of Soft Nanomaterials

UTILE-Gen: Automated Image Analysis in Nanoscience Using Synthetic Dataset Generator and Deep Learning

UTILE-Gen: Automated Image Analysis in Nanoscience Using Synthetic Dataset Generator and Deep Learning

Contact Info

Product

Resources

About