Replication Data for: Machine Learning for Improved Current Density Reconstruction from 2D Vector Magnetic Images (doi:10.7910/DVN/SD6PVP)

View:

Part 1: Document Description
Part 2: Study Description
Part 5: Other Study-Related Materials
Entire Codebook

Replication Data for: Machine Learning for Improved Current Density Reconstruction from 2D Vector Magnetic Images

doi:10.7910/DVN/SD6PVP

Harvard Dataverse

2025-01-13

2

Reed, Niko; Bhutto, Danyal; Turner, Matthew; Daly, Declan; Oliver, Sean; Tang, Jiashen; Olsson, Kevin; Ku, Mark; Rosen, Matthew; Walsworth, Ronald, 2025, "Replication Data for: Machine Learning for Improved Current Density Reconstruction from 2D Vector Magnetic Images", https://doi.org/10.7910/DVN/SD6PVP, Harvard Dataverse, V2

Replication Data for: Machine Learning for Improved Current Density Reconstruction from 2D Vector Magnetic Images

doi:10.7910/DVN/SD6PVP

Reed, Niko (University of Maryland)

Bhutto, Danyal (Boston University)

Turner, Matthew (University of Maryland)

Daly, Declan (University of Maryland)

Oliver, Sean (The MITRE Corporation)

Tang, Jiashen (University of Maryland)

Olsson, Kevin (University of Maryland)

Ku, Mark (University of Delaware)

Rosen, Matthew (Harvard University)

Walsworth, Ronald (University of Maryland)

Harvard Dataverse

Walsworth, Ronald

Reed, Niko

2025-01-10

https://doi.org/10.7910/DVN/SD6PVP

Computer and Information Science, Physics, Inverse Problems, Quantum Sensing, Machine Learning, Convolutional Neural Networks, Optically Detected Magnetic Resonance, Nitrogen vacancy centers in diamond

This repository contains the dataset and associated resources used in the scientific paper titled <b>Machine Learning for Improved Current Density Reconstruction from 2D Vector Magnetic Images</b>. The dataset includes: trained models ready for inference, validation data (including data shown in paper), experimental data (including data shown in paper), inference, data analysis, and data generation scripts, and Fourier inversion code for performance comparison.

Training Dataset 1: 50±50 μm standoff, 64x64 resolution: https://doi.org/10.7910/DVN/QPCS0I

Training Dataset 2: 500±50 μm standoff, 64x64 resolution: https://doi.org/10.7910/DVN/SJDS2O

Training Dataset 3: 50±50 μm standoff, 256x256 resolution: https://doi.org/10.7910/DVN/SJDS2O

Reed, N. R. et al. Machine Learning for Improved Current Density Reconstruction from 2D Vector Magnetic Images. Preprint at http://arxiv.org/abs/2407.14553 (2024).

arXiv:2407.14553

Reed, N. R. et al. Machine Learning for Improved Current Density Reconstruction from 2D Vector Magnetic Images. Preprint at http://arxiv.org/abs/2407.14553 (2024).

Label:

Article_Fulltext.pdf

Text:

Full text of article including supplemental information describing the performance of the MAGIC-UNet network on the magnetic inverse problem

application/pdf

Label:

ComparisonAndAnalysis.ipynb

Text:

Notebook containing inference function, Fourier inversion, basic plotting tools, and quantitative analysis tools. It can be run as a colab file if the repository is uploaded to google drive

application/x-ipynb+json

Label:

environment.yml

application/octet-stream

Label:

README.md

Text:

Read me! Contains information about dataset, instructions for inference, training, and data generation, license, acknowledgements.

text/markdown

Label:

requirements.txt

text/plain

Label:

64x64_close_standoff_experimentaldata.npy

Text:

Experimental data, various SNR and standoff distances (z<100 μm). 64x64 resolution

application/octet-stream

Label:

64x64_far_standoff_experimentaldata.npy

Text:

Experimental data, various SNR and standoff distances (z>100 μm). 64x64 resolution

application/octet-stream

Label:

validation_256x256_50um_Bxyz.npy

Text:

In distribution validation data (Input Bx, By, and Bz) for 256x256 resolution, standoff 50 ± 50 μm data

application/octet-stream

Label:

validation_256x256_50um_Jxy.npy

Text:

In distribution validation data (Ground truth Jx and Jy) for 256x256 resolution, standoff 50 ± 50 μm data

application/octet-stream

Label:

validation_64x64_500um_Bxyz.npy

Text:

In distribution validation data (Input Bx, By, and Bz) for 64x64 resolution, standoff 500 ± 50 μm data

application/octet-stream

Label:

validation_64x64_500um_Jxy.npy

Text:

In distribution validation data (Ground truth Jx and Jy) for 64x64 resolution, standoff 500 ± 50 μm data

application/octet-stream

Label:

validation_64x64_50um_Bxyz.npy

Text:

In distribution validation data (Input Bx, By, and Bz) for 64x64 resolution, standoff 50 ± 10 μm data

application/octet-stream

Label:

validation_64x64_50um_Jxy.npy

Text:

In distribution validation data (Ground truth Jx and Jy) for 64x64 resolution, standoff 50 ± 10 μm data

application/octet-stream

Label:

OutOfDistVal_64x64_50um_Bxyz.npy

Text:

Current density/magnetic field pairs generated entirely in COMSOL using different algorithm than training data

application/octet-stream

Label:

OutOfDistVal_64x64_50um_Jxy.npy

Text:

Current density/magnetic field pairs generated entirely in COMSOL using different algorithm than training data

application/octet-stream

Label:

blend_twotypes.py

text/x-python

Label:

file_converter.py

text/x-python

Label:

fix_problems.py

text/x-python

Label:

compressImages.m

text/x-matlab

Label:

deleteFiles.m

text/x-matlab

Label:

fgenerateRandomCurrent.m

text/x-matlab

Label:

fgenerateRandomGNDTERMLabel.m

text/x-matlab

Label:

fgenerateRandomSplineCoord.m

text/x-matlab

Label:

fgenerateRandomVoltage.m

text/x-matlab

Label:

fgenerateRandomWidth.m

text/x-matlab

Label:

fsetupCOMSOL.m

text/x-matlab

Label:

generate_current_densities.m

text/x-matlab

Label:

process_current_densities.m

text/x-matlab

Label:

readme.txt

text/plain

Label:

renameFiles.m

text/x-matlab

Label:

BinImage.m

text/x-matlab

Label:

generate_and_process.m

text/x-matlab

Label:

RightAngle2.m

text/x-matlab

Label:

Slope2.m

text/x-matlab

Label:

WideSlope2.m

text/x-matlab

Label:

keras_metadata.pb

application/octet-stream

Label:

saved_model.pb

Text:

Model trained with 64x64 resolution, standoff 500 ± 50 μm data

application/octet-stream

Label:

train_loss

Text:

Training loss recorded during training of 64x64 resolution, standoff 500 ± 50 μm network

text/plain; charset=US-ASCII

Label:

val_loss

Text:

Validation loss recorded during training of 64x64 resolution, standoff 500 ± 50 μm network

text/plain; charset=US-ASCII

Label:

variables.data-00000-of-00001

application/octet-stream

Label:

variables.index

application/octet-stream

Label:

keras_metadata.pb

application/octet-stream

Label:

saved_model.pb

Text:

Model trained with 64x64 resolution, standoff 50 ± 10 μm data

application/octet-stream

Label:

train_loss

Text:

Training loss recorded during training of 64x64 resolution, standoff 50 ± 10 μm network

text/plain; charset=US-ASCII

Label:

val_loss

Text:

Validation loss recorded during training of 64x64 resolution, standoff 50 ± 10 μm network

text/plain; charset=US-ASCII

Label:

variables.data-00000-of-00001

application/octet-stream

Label:

variables.index

application/octet-stream

Label:

unet_main_train.py

text/x-python

Label:

train.json

Text:

Config file to customize training parameters

application/json

Label:

unet_data_generator.py

text/x-python

Label:

unet_model.py

text/x-python

Label:

unet_trainer.py

text/x-python

Label:

config.py

text/x-python

Label:

dirs.py

text/x-python

Label:

utils.py

text/x-python