[Feature request] Possible to add support for Android?

[AndroidDeveloperLB]

Is your feature request related to a problem? Please describe. Yes. I'm interested in a solution for Android, to upscale images, but it seems very hard to find any solution of it.

Describe the solution you'd like Having a library that Android apps can easily use, to upscale images.

Describe alternatives you've considered Google offered only an example of using TensorFlowLite , but I can't find a way to create something myself, which uses a better model:

https://www.tensorflow.org/lite/examples/super_resolution/overview

https://github.com/tensorflow/tensorflow/blob/master/tensorflow/lite/g3doc/examples/super_resolution/overview.ipynb

Is it possible to add support for this repository on Android?

[Prurite]

Upscaling is a GPU intensive task and I think mobile devices can probably not provide the computational power required.

[AndroidDeveloperLB]

@Prurite Google provided a sample of it. Sadly of a tiny learning model that requires also a tiny image as input, but still... https://github.com/tensorflow/examples/tree/master/lite/examples/super_resolution Google uses something for its cameras too. https://research.google/blog/see-better-and-further-with-super-res-zoom-on-the-pixel-3/

[78Alpha]

That would end up CBS the entire program. For the conversion to tflite or the android pytorch, each model would need conversion and code changed to support the new tool chain. Otherwise, it would run in CPU mode and destroy battery life.

Usually the biggest issue is the model conversing. Need to know everything in it so it can be pulled apart and stitched back together. And if course, unexpected loss in accuracy.