BlindVision is an Android application designed to assist visually impaired users by providing real-time image classification and spoken feedback using the device’s camera and on-device AI.
It uses Jetpack Compose, CameraX, TensorFlow Lite, and TextToSpeech to recognize objects and communicate them aloud—creating an intelligent and accessible experience.
-
🎯 Real-time Image Classification
Detects and identifies objects using an on-device machine learning model. -
🧠 Confidence Threshold
Only announces results when accuracy is above 80%. -
🗣️ Speech Feedback (Text-to-Speech)
Reads out detected object and confidence for easy interaction. -
🖼️ HUD Overlay UI
Draws a sci-fi inspired visual HUD with label and confidence rating. -
🔘 Start/Stop Detection Toggle
Tap to start or stop image detection. Color-coded buttons for clarity.
| Technology | Description |
|---|---|
| 🛠️ Jetpack Compose | Declarative UI Framework |
| 🧪 TensorFlow Lite | On-device ML for image classification |
| 🧬 MediaPipe | Real-time camera frame handling |
| 📦 Hilt (DI) | Dependency Injection for ViewModels and services |
| 🧠 Kotlin | Primary language |
| 🧰 Android Studio | Official IDE for building the app |
- Android Studio Flamingo or later
- Minimum SDK: 23
- Internet: Not required
git clone https://github.com/yourusername/BlindVision.git
cd BlindVision