Tech Stack : Flutter, Tensorflow, OpenCV, Java, Swift, C#
This project was one of the most challenging one. Tensorflow Lite has implementations available on Android and iOS but no solid wrapper is available when it comes to Flutter. On top of that, running the model on real time frames was another technical challenge for us.
We successfully made the tflite model work by writing native implementations and then linking it to Flutter via platform channels. The image post processing to crop the relevant part was done by OpenCV’s native integration the same way. The project also required us to use C++ integration via dart:ffi to process the camera frames faster into a model readable byte buffer format.
With the experience working on this, we are now capable of running the same kind of native implementation for Tensorflow, Keras, OpenCV, CameraX or any other libraries via Flutter.
For years now, Adkrasol Design Studio was struggling to have a fully functional, navigable, fast loading, UI/UX rich and a mobile responsive website with all the advanced features.
No items found