Google Maps Accessibility Enhancer

• 

  google map

🚀 Inspiration

The inspiration came from observing the daily challenges faced by visually impaired people while commuting independently. I realised existing navigation apps do not provide real-time environmental awareness for such users, motivating me to build an AI-integrated accessibility enhancer.

🛠 How I Built It

I trained an object detection model using YOLOv5 and converted it to TensorFlow Lite for edge deployment on Android. The app was developed using Java and Android SDK, integrating Google Maps SDK for route data and Android’s Text-to-Speech and Vibrator APIs for voice and haptic feedback. OpenCV was used for camera processing and input normalisation.

📚 What I Learned

I learned how to optimise AI models for mobile, integrate multi-modal feedback mechanisms, and implement accessibility-focused UI/UX designs. I gained skills in combining AI inference pipelines with real-time voice guidance for enhanced situational awareness.

⚡ Challenges Faced

The major challenges were optimising the AI model to run smoothly on mid-range Android devices, ensuring low latency voice feedback, and designing haptic patterns that are intuitive yet non-distracting for users. Integration testing with accessibility services also posed challenges, which I overcame through iterative debugging and user feedback.

✏️ Math / Technical Note (LaTeX Example)

The object detection pipeline uses convolutional layers with an output tensor of shape $[1, N, 4+C]$ where $N$ is the number of detections and $C$ is the number of classes.

Built With

• android-accessibility-apis-tools:-android-studio
• github-hardware:-android-smartphone
• java-(android-sdk)-frameworks:-tensorflow-lite
• languages:-python
• opencv-apis:-google-maps-sdk
• optional