distance-transforms-paper

Structure

analysis.jl

Notebook that loads all of the various .csv files from the /data/* folder and prepares figures and tables for analysis and publication

benchmarks.ipynb

Google colab notebook that benchmarks our GPU accelerated python-wrapper distance transform against various Python-based distance transforms

benchmarks.jl

hausdorff_loss_timings.ipynb [In Progress]

Google colab notebook that benchmarks our DT in the hausdorff loss vs. other Python-based DTs vs. pure dice loss and benchmarks full training loops of the various hausdorff loss functions

hausdorff_loss_accuracy.ipynb [In Progress]

Google colab notebook that provides full training loop of pure dice vs hausdorff + dice loss accuracy comparisons
Outputs quantitative date (dice, hausdorff, iou, etc.) and qualitative images showing the different models

skeletonization.jl [TODO]

Notebook that implements a GPU based skeletonization algorithm (using our GPU distance transform) and compares against a pure CPU skeletonization algorithm (using a CPU distance transform)

/data

Contains the various timings and accuracy data as .csv files from the above notebooks

/plots

Name		Name	Last commit message	Last commit date
Latest commit History 82 Commits
.github/workflows		.github/workflows
data		data
plots		plots
.gitattributes		.gitattributes
.gitignore		.gitignore
LocalPreferences.toml		LocalPreferences.toml
Manifest.toml		Manifest.toml
Project.toml		Project.toml
README.md		README.md
analysis.jl		analysis.jl
benchmarks.ipynb		benchmarks.ipynb
benchmarks.jl		benchmarks.jl
hausdorff_loss_accuracy.ipynb		hausdorff_loss_accuracy.ipynb
hausdorff_loss_contours.ipynb		hausdorff_loss_contours.ipynb
hausdorff_loss_nnunet.ipynb		hausdorff_loss_nnunet.ipynb
hausdorff_loss_timings.ipynb		hausdorff_loss_timings.ipynb
skeletonization.jl		skeletonization.jl
visualizations backup 1.jl		visualizations backup 1.jl
visualizations.jl		visualizations.jl
visualizations2.jl		visualizations2.jl