search

kivy / kivy-ios

Toolchain for compiling Python / Kivy / other libraries for iOS
https://kivy.org/docs/guide/packaging-ios.html
MIT License
783 stars 241 forks source link

Support recipes for pandas, scikit-learn, tensorflow and h5py #649

Open mamunabcoder opened 3 years ago

mamunabcoder commented 3 years ago

Hi, I am trying to create Kivy-ios app to be able to view and analyse a data set. I see that python-for-ios doesn't have recipes for pandas, scikit-learn, tensorflow and h5py so I will have to create these myself. I have read the documentation throughly but didn't see any support of these library official support. Even I've read the article on how to create your own recipe but I don't know much about how to know if a module requires patches and what these patches should do, and what I may need to do in each of the four build methods.

I need support for these python library for iOS

Thanks in advance for any help.

RobertFlatt commented 3 years ago

Based on Android experience, Tensorflow is probably impractical for a mobile device. Tensorflow Lite has been compiled for all the desktop platforms (86 and arm) by others, and available on Android (via JNI), so that would be the best approach. But probably non-trivial. Porting a tf model to tfl is well documented.

SKL has not been ported to Android either.

guidocalvano commented 1 month ago

A single module capable of running onnx based models would make for some very cool demos. Think bounding boxes around objects on the camera. It would also open up a whole world of machine learning based apps running on the iphone through kivy-ios. https://onnxruntime.ai/docs/tutorials/mobile/deploy-ios.html With a little help I would look into it.