The words of this tutorial, step by step. The text lives in src/smappy/tutorial/topics/; the directions it follows are in src/smappy/tutorial/STYLE.md.
1. card: Simulating data: structures, blinking and camera frames
Data whose truth is known: to learn the program, to choose settings, and to test a fit.
One model gives localizations, or the camera frames to fit.
spoken: Simulated data has a known truth: to try the program, to choose settings, and to check a fit.
2. Simulate is in the File tab. What to make, how many frames, the background, drift and the random seed come first.
3. The structure: a built-in, here nuclear pores, or a file of your own.
4. card: A structure file
Label positions, or lines, circles, areas and images filled at a density; and how many copies, placed and turned at random.elements:
- points: [[0, 0, 0], [12, 5, 0]]
- line: [[0, 0, 0], [1000, 0, 0]]
density: 0.05
- circle: 50
density: 0.2
copies:
density: 2
rotation: random
spoken: A structure file lists label positions, or shapes filled at a density, and how many copies to place.
5. Not every site carries a dye: a labelling efficiency near 60 % is typical for the pores. Each label gets one fluorophore.
6. Blinking in the numbers a dye is known by: how long it is on, how often it comes back, and its photons per blink.
7. The off time follows from the number of blinks, and switching happens anywhere within a frame, as in a real movie.
8. Run, and the pores appear: eight corners each, some missing, as the labelling left them.
9. Two molecules on at once within a PSF cannot be told apart: they are removed, or averaged into one, as a fitter would see them.
10. Camera frames instead: the same molecules on a camera, with shot and read noise. They are saved as a recipe, not as a movie.
11. The recipe is a few lines: the settings and the seed. The frames are made as a fitter reads them.
12. Give the fitter the recipe as its file.
13. The camera comes from the simulation: nothing to type in.
14. Run.
15. Ground Truth works out where every spot really was and pairs them with the localizations, frame by frame.
16. Found, false and missed, and the Jaccard index that sums them up. Crowded spots and dim slivers cost detections.
17. With only the isolated spots, nearly all are found.
18. How many were found, by their photons: the dim ones are the hardest.
19. The real error against the precision the table reports. On each other: the precision can be trusted.
20. Each error over its own precision: a unit Gaussian when the fit is honest.
21. card: What to remember
- Structure: a built-in or a file of positions and shapes, copied many times.
- Labelling and blinking: efficiency, on time, blinks and photons, in the terms a dye is known by.
- Camera frames: a recipe the fitter opens, camera included.
- Ground Truth: what was found, and whether the precision is honest.
spoken: That is simulation: a known truth, to learn on and to test against.