To write more robust code that will work with many photos the parameters for opencv methods such as the HoughTransform need to be self-adjusting and not hard-coded. We'll have to:
Define the bounds of the params (the extreme values on either end)
Define conditions under which our detection is done or "good"
For example finding one pupil circle that's vaguely centered and in an area of high red density
Perform a binary or some sort of search on the space of parameters we could try
When a specific parameter set matches the conditions we've defined we're good.
To write more robust code that will work with many photos the parameters for opencv methods such as the HoughTransform need to be self-adjusting and not hard-coded. We'll have to: