An Android application that performs real-time door and obstacle detection using TensorFlow Lite and computer vision. Built upon TensorFlow's object detection example and Ultralytics' YOLOv11 implementation.
- Real-time object detection using multiple models:
- YOLOv11
- Hardware acceleration support (CPU, GPU, NNAPI)
- Audio feedback for detected objects
- Configurable detection parameters
- Optimized for accessibility and navigation assistance
- Android Studio Meerkat (2024.3.1) or newer
- Android device with SDK 24+ (Android 7.0+)
- Developer mode enabled on device
- Clone the repository
- Open project in Android Studio
- Sync Gradle files
- Build and run on physical Android device
The app allows customization of:
- Detection threshold
- Number of threads
- Maximum results
- Hardware delegate (CPU/GPU/NNAPI)
- Detection model
This project is dual-licensed under:
- Apache License 2.0 (TensorFlow components)
- GNU GPL (YOLO components)
Based on:
For detailed implementation explanation, see this Medium post.