Cocoa

Hi @josevalim, this should be ready for review now. :) I've added a precompilation guide for creating a precompiler module and using a precompiler. A demo precompiler module is available...

Hi @josevalim, thank you very much for reviewing the code. Here are some of my thoughts! > we can simplify the precompiler implementation if we move the logic for zipping/checksuming...

Okay, I think I got it! So in summary, we need to - have `precompile` return something like`{:ok, [{target, cache_dir}]}` then in `elixir_make` we go through the list, compress the...

I feel like the ImageDataGenerator could perhaps be a separate project and use Evision as a dependency (maybe an optional one). Potentially, it could also use other image processing backends...

Yeah, I think a library similar to the `imutils` could be the go. IMHO, although these image processing/manipulating operations/functions could belong here, it would be easier to maintain and more...

I'm closing this issue as I believe it would be better to implement this (and other related features) in another library.

Probably an upstream fix is needed as the Python script cannot correctly parse some C++ files in the OpenCV's 5.x branch. Tested on [2c2d7774](https://github.com/opencv/opencv/commit/2c2d7774b9bce2eaeb298f7726b813b42c0d9161).

Could be a separate library? Perhaps yes. But since nx is one of the dependencies of this project, it should be fine if I do it here.

To get an ROI of an image, we may use `Access` behaviour, the code will be somewhat similar to the following ```elixir img = Evision.imread!("image.jpg") img[{10..30, 10..30, :all}] ```

Checklist: ### Required callbacks - [x] constant - [x] from_binary - [x] eye - [x] iota - [x] random_uniform - [x] random_normal - [x] backend_deallocate - [x] backend_copy - [x]...