0

I want to build an android camera app with real time object recognition (similar to face recognition). I know that object recognition it's more complex than face recognition, so what I want to do is to use object-recognition algorithm for machine learning on a server and send from android to server frames that will pe processed by server and return back the object name that was found.

Right now the server receives a picture as parameter and response with the objects in the picture and I was thinking to use Camera.PreviewCallback and onPreviewFrame() to send the frame to server (deprecated in API level 21). Or should I do something with video streaming from android to server?

Can someone explain if it's possible and what it's the best way to do it?

Community
  • 1
  • 1
Jorj
  • 1,291
  • 1
  • 11
  • 32

1 Answers1

1

You can do the object recognition either on the device or, as you suggest, in the cloud - the advantage of the latter is the much greater processing power available, although this has to be balanced with the need to send the images to the server.

For both approaches there are some well known libraries or services which it would be worth looking at to see if they meet your needs.

For on device:

For server side it might be good to start by taking a look at Google's Cloud Vision service:

Mick
  • 24,231
  • 1
  • 54
  • 120