This sample implements an Activity that performs real-time object detection on the live camera frames. It performs the following operations:
- Initializes camera preview and image analysis frame streams using CameraX
- Loads a mobilenet quantized model using Tensorflow Lite
- Converts each incoming frame to the RGB colorspace and resizes it to 224x224 pixels
- 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.