Nice time to mark it as a milestone.
The repository has nicely documented working refactored code, examples, and visual explanation.
For now, it works only with round objects, such as the iris.
In the future, it would be nice to add elliptical object detection, and preprocessing steps.