Contact person: Offline Android Image Recognition

Phone:Show

Email:Show

Location: Coimbatore, India

Budget: Recommended by industry experts

Time to start: As soon as possible

Project description:
"I need an Android image-recognition app powered by TensorFlow Lite that works completely offline. Users should be able to point the phone’s camera at an object and immediately see what it is, even with airplane mode on. A clean way to update or swap the set of identifiable categories is essential, so I can expand or refine the model myself later.

Here’s what I’m after:

• Build the project in Android Studio using Kotlin or Java, integrating the TensorFlow Lite Interpreter for fast, on-device inference.
• Keep latency low and accuracy high on mid-range phones (Android 8.0+).
• Provide a straightforward mechanism—such as a labels file or model-swap routine—that lets me customise categories without touching core code.
• UI only needs to preview the camera feed, show the top prediction with confidence, and log results locally; no cloud calls or internet permission whatsoever.

Deliverables
1. Complete Android Studio project (source, gradle files, assets).
2. Optimised .tflite model with sample label set.
3. README explaining how to retrain/replace the model and update label files.
4. Demo APK or short screen-capture video proving offline detection works.

Acceptance criteria
• App installs and runs on real devices with no network dependency.
• Detection responds in under 200 ms per frame.
• I can swap in a new model/labels and see new categories reflected immediately.

Let me know which existing model or custom training pipeline you’d propose and your timeline to deliver." (client-provided description)