NavSight AI – Intelligent Vision-Based Navigation Assistant

Prototype Description
Cloud hosted frontend app page and Clean UI foor elderly people
NavSightAI is also accesible on Mobile and other devices
Precise prediction with high accuracy and confidence
Real-time detection of vehicles
Accurate human and people detection
Detection on Mobile with guidance

Inspiration

Navigation is something most people rely on daily without difficulty, but for visually impaired and elderly individuals, it can be challenging and unsafe. Existing navigation systems provide directions but often lack real-time awareness of the user's surroundings.

NavSight AI was built to bridge this gap by combining artificial intelligence with assistive technology. The goal is to create a system that not only guides users to their destination, but also helps them understand what is around them in real time, improving safety, confidence, and independence.

What it does

NavSight AI is an intelligent vision-based navigation assistant designed to help visually impaired and elderly individuals navigate safely and independently.

Allows users to input a destination and generates route guidance
Provides step-by-step navigation instructions through voice output
Uses a camera to detect obstacles such as people, vehicles, and objects in real time
Alerts users with safety messages such as "Obstacle ahead" or "Move left"
Combines navigation and computer vision to provide context-aware guidance
Prioritizes safety alerts over navigation instructions to prevent accidents
Continuously monitors surroundings for safer navigation

The project also includes a cloud-hosted frontend interface for improved accessibility and usability.

Built an interactive frontend using Streamlit
Cloud-hosted for easy remote access
Supports cross-platform usage on mobile devices, tablets including iPad, and PCs
Allows users to access the system directly through a browser without complex installation

How we built it

Developed a real-time system combining navigation and object detection simultaneously
Integrated camera input, AI models, APIs, and voice systems
Designed a decision-making system to prioritize obstacle alerts over navigation instructions
Optimized detection performance for low latency and accuracy
Built and deployed a frontend interface using Streamlit for browser-based interaction

Challenges we ran into

Running object detection and navigation together in real time without delay
Integrating multiple components including AI models, APIs, frontend, and voice systems
Designing a smart alert system that is useful without overwhelming the user
Balancing detection accuracy with performance constraints
Ensuring cloud deployment and cross-platform compatibility across devices

Tech used

Python
Ultralytics YOLO
OpenCV
Streamlit
pyttsx3 / gTTS
Cloud hosting platform

What we learned

Integrating AI, APIs, cloud deployment, and frontend systems into one project
Building real-time object detection systems with practical latency constraints
Designing safety-first decision-making systems
Creating accessible interfaces for assistive technology
Deploying applications for browser-based cross-platform support
The importance of real-world testing and iterative improvements

What's next for NavSight AI – Intelligent Vision-Based Navigation Assistant

Implement more accurate distance estimation using depth sensing or stereo vision
Add indoor navigation support where GPS is unavailable
Develop a wearable version such as smart glasses or a compact device
Build a dedicated mobile application
Improve AI models for better recognition in complex environments
Add real-time rerouting based on dynamic obstacles and conditions
Integrate cloud features for route optimization and analytics