We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Research data for Microplastic quantification in food samples, using stereomicroscopy and smartphone camera, leveraging interactive supervised machine learning tools - JPCE
AI summary Read the abstract
Researchers developed and published an image analysis pipeline using stereomicroscopy and smartphone cameras combined with interactive machine learning tools to quantify microplastics in food samples, releasing raw images, trained classifier files, and processing macros to enable reproducibility of the workflow.
Research data for the article "Microplastic quantification in food samples, using stereomicroscopy and smartphone camera, leveraging interactive supervised machine learning tools". This repository contains the datasets and processing files used in the microplastic detection workflow developed in this study. The files included allow reproducibility of the image analysis pipeline. The repository contains: (i) raw microscopy images (.png) corresponding to the validation dataset used to evaluate the trained models; (ii) trained Ilastik project files (.ilp) for the two classifiers developed in this work (filament and fragment detection), which can be opened directly in the Ilastik software* to reproduce the image probability segmentation step; and (iii) FIJI macros (.ijm) used for image preprocessing and automated particle analysis in Fiji.
More Papers Like This
Research data for Microplastic quantification in food samples, using stereomicroscopy and smartphone camera, leveraging interactive supervised machine learning tools - JPCE
AI summary Read the abstract
Researchers released a dataset and image analysis pipeline — using stereomicroscopy, smartphone imaging, and supervised machine learning — for detecting and quantifying microplastics in food samples, enabling reproducible filament and fragment classification without specialized laboratory equipment.
Research data for Microplastic quantification and chemical characterization in salt samples, using stereomicroscopy, smartphone camera and supervised machine learning tools.
AI summary Read the abstract
Scientists developed a new, low-cost method to detect and count tiny plastic bits (microplastics) hiding in salt using a smartphone camera, a basic microscope, and smart computer software instead of expensive lab equipment. This matters because salt is something almost everyone eats daily, and having a cheaper, more accessible way to check it for plastic contamination could help researchers and food safety agencies track this potential health concern more widely. Note that this particular resource is the underlying dataset and code shared for other scientists to verify and reuse the method, rather than a report of new findings about how much microplastic is actually in salt.
A Handy Open-Source Application Based on Computer Vision and Machine Learning Algorithms to Count and Classify Microplastics
AI summary Read the abstract
An open-source computer vision application was developed to automatically count and classify microplastics in microscopy images, achieving accuracy comparable to manual counting while processing samples orders of magnitude faster, offering the scientific community a free tool to reduce the bottleneck of tedious visual microplastic enumeration.
A new approach for routine and accelerated processing of microplastics using MPScanTool plugin for ImageJ
AI summary Read the abstract
Researchers developed a flatbed scanner-based system paired with open-source image analysis software to speed up the counting, measuring, and classification of microplastics in samples, reducing the time required compared to manual microscopy. The affordable system makes high-throughput microplastic analysis more accessible to laboratories with limited resources.
Deep learning based classification of microplastic in edible food using optical microscopy images
AI summary Read the abstract
Researchers built an image dataset of microplastics extracted from edible food and benchmarked six deep learning architectures for classifying particles into threads, beads, and fragments, with MobileNetV2 and ResNet variants achieving 98% accuracy — demonstrating AI-driven classification as a scalable alternative to labor-intensive manual identification.
Research digests by email
When a large batch of papers lands in the Atlas, we read through it and send a short write-up of what stood out.