VisionGuard

Empowering the Blind with Smart Mobility and Real-Time Assistance

Comment

Inspiration VisionGuard was inspired by the idea of combining real-time communication, smart assistance, object detection, and geolocation to empower visually impaired individuals. Our goal was to create a tool that connects blind users to human assistants while also providing AI-powered help, real-time location sharing, and enhanced environmental awareness through object detection. We wanted to build a support system that is always available — whether through a trusted contact, an AI agent, or a shared live map.

What it does VisionGuard is a mobile application designed to assist blind users in multiple ways:

Real-time video calls with a human assistant (family member, volunteer, etc.).

Live geolocation tracking, so the assistant can see where the blind user is in real time on a map.

An AI Agent that helps the user by performing useful tasks such as web searches or sending pre-written emails — for example, in case of emergencies or when no assistant is available.

Object detection using YOLOv4, which helps identify surrounding objects in real time, enhancing spatial awareness and safety for visually impaired users.

Together, these features create a smart and flexible support system for blind users to get help when they need it.

How we built it The app was built in Flutter and divided into several integrated modules:

The video communication module uses Agora SDK to enable live video and audio between the blind user and the assistant.

Supabase manages real-time messaging (incoming call, call accepted, call ended), user authentication, and backend services.

The geolocation module relies on Google Maps and Flutter plugins like Geolocator to share the user's location in real-time.

The AI Agent is implemented using a smart automation system with n8n, handling tasks like internet searches or email sending based on voice/text input.

The object detection module integrates YOLOv4, which processes real-time camera input to recognize objects in the user’s environment and provide audio feedback.

Each member of the team contributed to one or more modules, ensuring tight integration and a seamless user experience.

Challenges we ran into

Managing camera, microphone, and location permissions across Android and iOS.

Handling asynchronous messaging and real-time updates using Supabase.

Ensuring accurate, responsive, and battery-efficient GPS tracking.

Designing the AI Agent to be helpful while remaining private and simple.

Implementing real-time object detection on mobile devices with acceptable performance.

Creating an intuitive, accessible UI for blind and low-vision users.

Accomplishments that we're proud of

Established seamless real-time video communication with low latency.

Integrated a live map that continuously updates the user's location for assistants.

Developed an AI Agent capable of assisting through automated tasks.

Successfully integrated YOLOv4 for real-time object detection, enhancing the user’s environmental understanding.

Built a modular and scalable architecture ready for future improvements.

Delivered a functional prototype meeting key accessibility needs.

What we learned

Advanced integration of services like Agora, Supabase, Google Maps, and YOLOv4 within Flutter.

Managing real-time data synchronization and communication flows.

Designing for accessibility-first use cases, especially for visually impaired users.

Implementing object detection on resource-constrained mobile devices.

Importance of modular design and effective team collaboration.

What’s next for VisionGuard

Add voice control for fully hands-free interaction.

Expand the AI Agent with automation like booking appointments or translating text.

Enhance geolocation with predictive movement and safety zones.

Build a community network of volunteer assistants within the app.

Enable offline emergency functionality with preloaded messages and contacts.

Improve object detection accuracy and expand detection categories using advanced YOLOv4 models.

Built with:

Flutter – for cross-platform mobile development

Dart – primary programming language

Agora SDK – for live video/audio communication

Supabase – for authentication, real-time database, and signaling

OpenStreetMap – for location tracking

Geolocator – for accessing and updating GPS data

AI Agent with n8n – for automation tasks (email, search)

YOLOv4 – for real-time object detection

Provider / ViewModel – for state management

QR Code libraries – for secure data sharing

Permission Handler – for managing hardware access

Android & iOS – deployment platforms

Built With

Share this project:

Updates