Skip to content

A comprehensive Android application package (APK) solution for on-device MLOps

License

Notifications You must be signed in to change notification settings

nnstreamer/nnstreamer-android

Repository files navigation

A Single APK integration of NNStreamer and ML API

Dokka

This project offers a single Android application package (APK) solution for on-device Machine Learning Operations (MLOps). As one of the core components, this APK includes an Android Service that uses ML API to maintain NNStreamer pipelines and Machine Learning (ML) models for those pipelines. The Service exposes multiple end-points to address ML computation offloading requests from external Tizen and Android devices, as well as ML task delegation requests from other Android applications running on the same device. The APK also provides an Android Activity that controls the Service's features. This Activity enables users to supervise the ML models and pipelines while hosting a user interface to visualize the results of the ML computations. In addition, sample applications are provided as concrete examples to demonstrate how to use the Service to perform external ML computation offloading and internal delegation of ML task requests.

Documentation

  • KDoc documentation generated from Kotlin source code is available at here.

Use-Case Scenarios

Tizen-to-Android ML Offloading

Component Interaction Diagram

Component Interaction Diagram for the Tizen-to-Android Scenario

Disclaimer

Note that the license of this repository, Apache-2.0, is only valid for the set of Android components provided this repository. Other software components used in this project, such as: NNStreamer, ML API, TensorFlow Lite, etc. (especially, the files and the cloned repositories under the externals directory), are provided as-is and are governed by their respective licenses.

About

A comprehensive Android application package (APK) solution for on-device MLOps

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages