Currently, the two TFLite models for FaceNet and FaceNet-512 are stored in the assets folder of the project and are bundled with the APK. This increases the size of the APK greatly.
Instead, the models can be stored in a GitHub release of this repository, and downloaded at runtime (after installation) from the app using the Ketch library. The models will then be downloaded in the app's internal storage (filesDir) and accessed by the TFLite Interpreter for inference.