google-ai-edge
Request: offer image-enlarging/unscaling/enhancing library that uses AI
I was told to create this request from here: https://github.com/google-ai-edge/LiteRT/issues/182#issuecomment-2691726472
MediaPipe Solution (you are using)
none
Programming language
Android Java
Are you willing to contribute it
None
Describe the feature and the current behaviour/state
All I've found are either very complicated, or very restricted, or require a server as it's too much work done locally
Will this change the current API? How?
No response
Who will benefit with this feature?
No response
Please specify the use cases for this feature
Every app that wishes to show/create a bit better image quality of given images, especially low quality or blurry ones
Any Other info
Please offer a very easy-to-use API, that I provide a Bitmap instance, a target resolution (can of any size, including larger), and I get an upscaled/enhanced version of it. You can also add flags such as requesting to add color in case it detects that it's black&white, for example.
Please also offer various trained models to use easily, and a tutorial of how to choose one of them, so that the extra size of the app would be our choice, as models can be quite large... Maybe some of them would be more focused on certain types of content.
We are currently not planning such a capability/pipeline in MediaPipe, and if anything changes we'll comment here.
Meanwhile, MediaPipe offers an ImageToTensorCalculator + InferenceCalculator which you can try to use along with your custom model to create a custom pipeline to upscale the image.
@whhone Thank you. I don't know how to use any of these, let alone together. In the beginning we had TensorFlow Lite, then we got Lite-RT, and now we have MediaPipe? I can't keep up even with the names...