Face-X is a modular and high-performance face recognition system built for Android. It supports multiple face detection and recognition models, allowing seamless switching between OpenCV, MediaPipe, ML Kit, and LiteRT.
The app is optimized for real-time inference, featuring low-latency processing, dynamic model selection, performance analytics, and interactive UI effects.
✅ Supports OpenCV, MediaPipe, ML Kit, and LiteRT for detection and recognition.
✅ Dynamic model switching for benchmarking performance.
✅ Achieves real-time inference with GPU acceleration (~10ms recognition, ~15ms detection).
latecny.mp4
👉 My detection (orange box) used in the test video above.
will-chad.mp4
"Chad Smith looks 82% similar to Will Ferrell before increasing the threshold!"
famhy_3.mp4
"Hussein Fahmy and his brother Mustafa Fahmy, Egyptian actors, famously resemble each other."
✅ Displays live processing times for face detection & recognition.
✅ Dynamic line chart updates in real-time.
✅ Skip, pinpoint, and data recording (up to 5 minutes of tracking).
linecart.mp4
✅ Unique disintegration & assembly effects for any Jetpack Compose UI component.
✅ Smooth animation transitions for interactive UI elements.
particles.mp4
✅ Enables a fixed footer in the bottom sheet for persistent controls.
✅ Enhances usability and quick access to settings or actions.
pinnedFooter.mp4
icons.mp4
"I chose the old TensorFlow logo because I got the idea to extract the T-F-L letters as it works on TensorFlow Lite. This was the hardest logo to design in Compose."
"The OpenCV animation is not very good. I had another idea, but it would take more time, and I have a lot of things to finish in the app."
We're continuously improving Face-X to provide more flexibility, control, and performance optimizations. Here’s what’s coming next:
✅ Ability to add and configure your own face detection & recognition models.
✅ Support for more model architectures, including high-performance quantized models.
✅ Automatic model benchmarking to suggest the best model for your device.
✅ More hardware control, including manual selection of CPU, GPU, and NPU.
✅ Advanced device analytics, including memory usage, battery impact, and thermal throttling insights.
✅ Optimized low-power mode for extended usage on mobile devices.
✅ Ability to prioritize performance or efficiency based on user preferences.
✅ Adaptive processing to dynamically adjust processing power based on workload.
✅ Background processing mode to allow passive face recognition while running other tasks.
Stay tuned for these updates! 🚀
Face-X is an ongoing experimental project where I continuously challenge myself to explore and refine ** Machine Learning integration and performance optimization techniques**.
✅ A platform for innovation – Experimenting with cutting-edge ML models, hardware acceleration, and real-time image processing.
✅ A continuous learning journey – Exploring new architectures, optimizing inference performance, and refining system efficiency.
✅ Future-focused – Expanding capabilities with custom model support, enhanced hardware controls, and advanced system monitoring (battery, memory, CPU usage).
This is not a finalized product but rather a dynamic and evolving initiative. If you're interested in collaboration, discussion, or feedback, feel free to connect! 🚀
Face-X is currently a personal demo project developed as a practice-driven exploration of modern face recognition technologies.
🔸 Not production-ready – This project is a sandbox for testing models, architectures, and optimizations.
🔸 Actively evolving – Features, performance metrics, and system integrations are subject to change as development progresses.
🔸 Independently developed – Built as a solo initiative for technical growth and experimentation.
I'm always open to discussions, ideas, and collaborations. Feel free to reach out if you're interested! 🚀
- Languages: Kotlin
- UI: Jetpack Compose
- Camera & Image Processing: CameraX
- Machine Learning Integration with Android: TensorFlow Lite (LiteRT), OpenCV, MediaPipe, ML Kit
- Architecture: Clean Architecture,MVVM
- Concurrency: Coroutines, Flow