Skip to content

Latest commit

 

History

History
 
 

CameraXTfLite

CameraX + Tensorflow Lite

This sample implements an Activity that performs real-time object detection on the live camera frames. It performs the following operations:

  1. Initializes camera preview and image analysis frame streams using CameraX
  2. Loads a mobilenet quantized model using Tensorflow Lite
  3. Performs inference on the transformed frames and reports the object predicted on the screen

The whole pipeline is able to maintain 30 FPS on a Pixel 3 XL with:

  • the default image size from Camera (640x480)
  • the default tensor size (300 x 300)

Screenshots

demo screenshot 1