I’m building an app that can recognise banknotes through the phone’s camera on both iOS and Android. The core flow is simple: the user points the camera at a note, the app instantly identifies the denomination and returns the result on-screen (voice feedback would provide). My immediate priorities • Real-time camera detection that works reliably in varied lighting and angles. • A companion “dataset mode” where I can feed new note images, label them, and run quick tests inside the app to verify model accuracy before pushing an update. • Initial coverage must include Pakistani Rupee notes; I’m open to extending to other currencies once the pipeline is proven. Model You are required to use different types (at least 3 models) of models with different types (3-4 types) of datasets after building the application. Finally, you should select the best model that provides the best and maximum accuracy. Tech flexibility I’m fine with Flutter, React Native or a pure native split—choose whichever lets you ship a smooth cross-platform build and integrate an on-device inference engine such as TensorFlow Lite, Core ML or ML Kit. Model size and speed should remain reasonable for mid-range phones. Deliverables 1. Complete source code and project files for iOS and Android. 2. All models and all datasets 2. Trained model (or training pipeline) with documented steps to add new currencies. 3. A short read-me and screen-recorded demo showing real-time detection and the internal dataset test feature. NB: I will provide three files, file 1: comments that you must follow and fix/build the project based on this file comment. File 2: that is main project where you will find all details regarding the app, programming, model and dataset. File 3: this is methodlogy and you will follow and update based on comments from file 1. If this sounds like your wheelhouse, let’s talk timeline and milestones.