Talk title: Reproducible Segmentation of Not-Quite-Objects
Presentation materials for my talk on cilia segmentation for Jupyter Day ATL.
The main content is in the ipynb notebook file.
The data needed by the analyses in the notebook are found in the data folder, which has two subfolders:
videos: threenpyfiles that contain the grayscale videossegmaps: threenpyfiles that contain the black-and-white ground-truth segmentation masks for each video
You can load any of these data yourself using the numpy.load function.
The other folder is spq, which is a module containing some helper code for the analyses in the notebook.
spq.widgetshas the full code needed for generating the variance-based threshold widgetsspq.utilshas the parameter-scan versions of the same functionsspq.evaluatehas the function that implements intersection-over-union for evaluating our predicted masks against the ground-truth
The easiest way is to click the "launch binder" button at the top of this README. That will launch a new tab in your window and spin up this repository as an active Jupyter environment. It may take a few minutes but you should be able to re-run everything in the notebook, even editing to your heart's content.
If you want greater control, you can clone this notebook, make sure you have the prereqs installed (via environment.yml), and then simply jupyter notebook your way to success.
- Tweet at me @SpectralFilter
- Email me spq@uga.edu
- Open up a ticket here, I guess?
- Smoke signals are cool too
MIT