This repository hosts the code and related resources for the 2DMatGMM project, a robust machine learning system designed to detect and classify 2D material flakes in real-time. The details of the project are published on arXiv in the paper titled "An open-source robust machine learning platform for real-time detection and classification of 2D material flakes".
The paper comprises three repositories each containing a part of the project:
This Repository contains the dataset used in the paper. The dataset will be made available for download on Zenodo. It is currently privately hosted. The download links as well as metadata about the dataset and usage can be found in the Dataset Guide.
To install and set up your system to use 2DMatGMM, follow the installation instructions.
This repository comes with weights pretrained on the datasets used in the paper. To use the detector for you own projects check out the Getting Started guide.
You may also quickly hop in by trying out the demo on Google Colab. The demo is a Jupyter Notebook which runs the detector on a few demo images from the testset of the Graphene Dataset.
If you use our work or dataset in your research or find the code helpful, we would appreciate a citation to the original paper:
Jan-Lucas Uslu et al 2024 Mach. Learn.: Sci. Technol. 5 015027
@article{Uslu_2024,
doi = {10.1088/2632-2153/ad2287},
url = {https://dx.doi.org/10.1088/2632-2153/ad2287},
year = {2024},
month = {feb},
publisher = {IOP Publishing},
volume = {5},
number = {1},
pages = {015027},
author = {Jan-Lucas Uslu and Taoufiq Ouaj and David Tebbe and Alexey Nekrasov and Jo Henri Bertram and Marc Schütte and Kenji Watanabe and Takashi Taniguchi and Bernd Beschoten and Lutz Waldecker and Christoph Stampfer},
title = {An open-source robust machine learning platform for real-time detection and classification of 2D material flakes},
journal = {Machine Learning: Science and Technology},
}
If you encounter any issues or have questions about the project, feel free to open an issue on our GitHub repository. This Repo is currently maintained by Jan-Lucas Uslu.