Published April 20, 2024
| Version v2
Dataset
Open
MOKE-microscopy-Skyrmion-dataset
Authors/Creators
- 1. Johannes Gutenberg-Universität Mainz, Germany
- 2. Max-Planck-Institut für Struktur und Dynamik der Materie Hamburg, Germany
-
3.
Johannes Gutenberg University Mainz
Description
Magnetic skyrmions in thin-film systems are heavily under investigation due to their potential applicability for spintronic applications.
An often used imagning techniqe is the magneto-optical Kerr effect (MOKE-)microscopy. Here we provide approximately 2400 MOKE images of around 30 different thin film measurements, focused on magnetic skyrmions and material defects. The dataset counts more than 1 million skyrmions. Corresponding labels are also provided pixelwise as a 3-class label (Background, skyrmion, defect) for training neural network devices on detection.
Own U-Net based pre-trained models are available for direct usage.
The corresponding publication is in preparation. Statistics and models used for publication are also included.
Notes
We thank all the experimentalists in the Kläui-Lab that took the MOKE measurements over the past for their own projects.
The data is split in fixed training/validation and test set, as the training labels are considered to be relatively weak - due to some relabeling process -, while the validation and test set are labeled carefully by hand.
Files
public_unet_skyrmion_dataset.zip
Additional details
Related works
- Continues
- Dataset: https://zenodo.org/doi/10.5281/zenodo.7636109 (URL)
- Journal article: 10.1103/PhysRevApplied.21.014014 (DOI)