Android CameraX image analysis demo with OpenCV4 and TesorFlow Lite
(Work in progress) Develop virtual camara mode.
Real camera mode
[Image sensor]-->[CameraX]-->[OpenCV4]--+-----------------------+
| | |
+------------------+--->[TensorFlow Lite]--+---> Final output
Virtual camera mode (for Unity)
[Unity camera]-->[RenderTexture]-- MQTT --->[OpenCV4]--+-----------------------+
| | |
+------------------+--->[TensorFlow Lite]--+---> Final output
I think 4G/5G smart phones (or 4G/5G with Android-based cameras) with 4K/8K image sensors will replace the traditional PTZ monitoring cameras.
Traditional PTZ monitoring camera
Edge AI
[Camera]---ONVIF/PoE---[Edge computer]---Ethernet---[Video recorder]---[Video management system]
Android-based monitoring cameras are cheaper and more flexible than the traditional ones.
Edge AI
[Smartphone]---4G/5G network---[Cloud storage]---[Video management app on cloud]
|
Direct communication over 4G/5G network
|
[Smartphone]
This project is just a skeleton of such an AI camera.
OpenCV4 Android SDK has to be installed properly in this project -- it is not included in this repo.
The other components are installed by Gradle automatically.
- Color filter
- Optical flow
- Difference extraction
- Object detection