Camera-Machine Learning Environment Research (CaMLER) Project

Introductions

Nowadays, machine learning library such as TensorFlow has evolve rapidly so it can enabled of mobile or smartphone doing their own machine learning tasks independently. However, due to the complexities, not all of (machine learning) models can run in smartphone. Instead of doing model simplification so it can run in smartphone with the risk in reducing the quality of the outputs, we will run the model in the server (cloud). In this research project, we will create an environment research where client (application) can send the image classification request to the server and the server will doing the classification and sent back the result to the client.

Client

Introductions

Instead of deploying the model in the client itself, we will create an application (client) that have ability to take the picture and send the request to the server and wait for the result or feedback.

Constraints

• Android application, native, written in Kotlin.
• Minimum Android SDK 21 (5.0 a.k.a Jelly Bean)
• Compiled and target SDK is 29
• Image captured should be saved in media folder
  

  /Android/media/domain.applicationdomainname.applicationsubdomainname/applicationname/images/
  

• Image filename should written in format of yyyy-mm-dd hh:mm:ss and saved in format of *.jpg
• Canceled image should be removed from the media folder
• Allowed dependencies,
  
  • Android standard dependencies
  • Android JetPack dependency (CameraX)
  • Android Kotlin Coroutines library
  • Google Materials
  • Image: Picasso or Glide
  • Network: GSon, KTOR, OkHTTP
  • Asynchronous: ReactiveX (rxJava, rxKotlin)
• Please use the latest version for all dependencies that used in this project
• Dependencies other than in those list should have discussed first

Updates

• 2020/06/22 13:10 WIB
  
  • Only used Android JetPack CameraX dependency for camera library.
  • If the user exit or destroy the application before the result is retrieved, bring back the Result Activity if user open up the application again and re-fetching the request.
  • Remove the image file from the media folder if the result has been retrieved.

Server

To be discussed later.

Contributors

• PIC
  Nandhika Prayoga
• Supervisor
  Arief Purnama Muharram