Christoph Rackwitz

Would it be too unreasonable to stick as closely as possible to esp-idf APIs?

My bad. The name made me think this was using espressif APIs. I didn't realize this was actually based on Arduino.

the Stack Overflow questions that those people came from: * https://stackoverflow.com/questions/77897622/opencv-add-operation-results-in-unexpected-dtype-and-shape * https://stackoverflow.com/questions/78111233/strange-behaviour-of-cv2-add-operation I'd say it's **not** a bug but an unexpected feature that trips newbies up because they aren't...

I'd recommend using `.reshape()`... ... and to *never ever* say "double bracket". That wording misses the point (three dimensions) and may misguide newbies.

You know what I want? To annotate local variables/assignments: ```python def azimuth_elevation_from_equirect(u: f4, v: f4) -> typing.Tuple[f4, f4]: pano_cx: f4 = (pano_width-1) * 0.5 pano_cy: f4 = (pano_height-1) * 0.5...

I really wish the docs had a specific section explaining specifically how I can be absolutely sure (by telling numba) which things are supposed to be which types, especially when...

Something *like* this [just popped up on Stack Overflow](https://stackoverflow.com/questions/79631895/why-does-the-variance-result-exceed-0-when-using-the-tm-sqdiff-method-to-match-a), with the difference that `cv.setUseOptimized()` seems to have no effect and either True or False result in a wholly non-zero `matchTemplate`...

I must also question the utility of a simple "best" attribute. This merely splits available imagery sources into "first class" and "second class", without quantitative meaning. This is especially noticeable...

Would be good to see that documented explicitly, both next to the method's API doc and in whatever examples there are.

The point of remap is to work with such lookup tables. Some mappings cannot be represented other than with lookup tables. if you need something that is different from remap(),...