Particle Classification

Welcome to the help and tutorial documentation for the ParticleTrieur program and the MISO python library.

Overview

MISO

MISO is a library of python scripts that simplify training a CNN from a set of labeled images. A variety of common CNN topologies can be chosen, such as variations of ResNet or using transfer learning. The scripts take a folder of images and output a trained model along with statistics on the model performance. The system is optimised for particle images.

Github repository

ParticleTrieur

ParticleTrieur is a cross-platform java program to help organise, label, process and classify images, particularly for particle samples such as microfossils. It can be used for both the creation of the training set required to make a CNN classifier, and classification of image using a trained CNN. It also includes some image processing functions, morphology calculations and statistical graph generation. ParticleTrieur allows the user to configure and launch training using the MISO library.

ParticleTrier is release under the open-source GPL v2 licence and the source code can be found at Github repository

Getting started

Read the introduction on ParticleTrieur, MISO and how to create a good training set for CNN models.
Install ParticleTrieur and MISO

Citing

If using ParticleTrieur with the MISO library please cite our paper on CNNs for foraminifera classification at the Journal of Micropalaeontology

@article{jm-39-183-2020,
   author = {Marchant, R and Tetard, M and Pratiwi, A and Adebayo, M and de Garidel-Thoron, T},
   doi = {10.5194/jm-39-183-2020},
   journal = {Journal of Micropalaeontology},
   number = {2},
   pages = {183--202},
   title = {{Automated analysis of foraminifera fossil records by image classification using a convolutional neural network}},
   url = {https://jm.copernicus.org/articles/39/183/2020/},
   volume = {39},
   year = {2020}
}

Table of contents

Getting started

ParticleTrieur