Hazards slip through
A cane finds what’s on the ground, but can miss overhead hazards and temporary barriers like chest-height construction signs.
Experimental AI prototype · Accessibility research
See the World
Differently.
Vision Assist uses a phone camera and AI to spot what is around you — people, cars, bikes and more — and tells you out loud, so you can keep walking with more confidence.
- Detects
- 80object types
- Alerts
- Voicewith direction
- Privacy
- On-devicein this demo
YOLOv8n · live analysis4 objects
Spoken alert
“Caution. Bus ahead, close. Bus on your left, close. Person ahead, close.”
- 2.2 billion
- people worldwide have a near or distance vision impairment.Source: World Health Organization, 2026 (opens in a new tab)
- 2 million+
- people in the UK are living with sight loss that affects their daily lives.Source: RNIB (opens in a new tab)
- 250
- people in the UK start to lose their sight every day — one every six minutes.Source: RNIB (opens in a new tab)
- 0
- extra devices needed. Vision Assist runs on the smartphone people already carry.Vision Assist design goal
The problem
Getting around shouldn’t mean stopping to scan.
Through research, accessibility reports and conversations with potential users, the same issues came up again and again.
Constant concentration
Walking somewhere unfamiliar takes intense focus. Over a whole journey, that mental load becomes exhausting.
Less confidence, fewer trips
When every outing is hard work, people go out less often — which can lead to isolation and lost independence.
Why today’s tools fall short
Navigation apps
Great at routes and arrival times.
They miss the last few metres: bollards, street furniture and temporary barriers.
Scene-description apps
Excellent at reading text and describing a scene.
You usually stop and take a photo, which breaks the flow of walking.
Wearable devices
Hands-free and powerful.
Often cost £500 or more and mean carrying extra gear.
The gap: continuous, walking-level hazard awareness — on the phone you already own.
How it works
Three steps. No screen needed.
Step 1: Hold or wear your phone
Point the camera ahead, at chest height or on a lanyard. Start Vision Assist with your voice.
Step 2: AI watches the path
Every frame is checked for people, vehicles and obstacles, with their direction and rough distance.
Step 3: Hear what matters
Short spoken alerts — “Person at front left”, “Car right, 10 metres” — only when something matters.
What it does
Built for the way you actually move.
Every feature is designed around one idea: tell people what matters, when it matters, without slowing them down.
Hazard alerts while you walk
No stopping to take photos. Vision Assist keeps watching and speaks up when something is in your way.
Direction and distance
Alerts say where something is — left, ahead or right — and roughly how far, from object size and depth estimation.
Voice first, screen optional
Start it, ask for directions and hear alerts without looking at the screen. Works with screen readers.
Works with your cane or guide dog
Designed to add warnings your usual aid can’t give, like chest-height signs — not to replace it.
No special hardware
Runs on a standard modern smartphone. No glasses, sensors or subscriptions to start.
Private by design
Video is processed on the device wherever possible. Camera footage is not uploaded or stored.
Real-world test
Tested on real streets, not just in the lab.
Field test: walking to the University of Bristol
Real-world testThe app is started by voice, asked to navigate to the University of Bristol, and gives spoken directions and hazard alerts along the way. The phone screen on the right shows each alert as it is spoken.
Read the transcript
- 0:00“Activate.” The app confirms: scan activated.
- 0:04“Navigate to University of Bristol.” — “Navigating to University of Bristol.”
- 0:08“Head southeast toward Old Market Roundabout, 38 metres.”
- 0:12“Head north toward Wine Street, 4 metres.”
- 0:18“Blind road at front left.” (tactile paving ahead on the left)
- 0:22“Person at front.”
- 0:24“Head northwest on Broad Street toward Corn Street, 0.2 kilometres.”
- 0:28“Person at front left.” — repeated while walking behind other pedestrians.
- 0:42“Blind road at front right.”
- 0:46“Speeding car at front right, 10 metres away.”
- 0:50“Head southeast toward Tyndall Avenue, 10 metres.”
In this one-minute walk through Bristol, Vision Assist is started by voice, asked for directions to the University of Bristol, and warns about what’s ahead on the way.
- Started and controlled entirely by voice
- Turn-by-turn directions with distances
- Alerts for people, with their direction
- Tactile paving (“blind road”) noticed along the route
- A fast-moving car flagged at 10 metres
Live demo
Try the AI in your browser.
Upload a street photo and hear what the model finds. It runs on your device — your image is never uploaded.
- Real YOLOv8 object detection
- Spoken alerts with direction
- Works with keyboard and screen readers
Who it’s for
Built with the people who will use it.
People with sight loss
Blind and partially sighted adults who use a smartphone and want more confidence on everyday walks.
Mobility specialists & charities
Habilitation and rehabilitation specialists and sight-loss charities who can help shape and test it.
Universities, transport & venues
Organisations that want accessible campuses, stations and buildings for blind and partially sighted visitors.
Roadmap
Where we are, and where we’re going.
- DoneDone
Research & prototype
- Problem research and market validation
- Technical feasibility and MVP architecture
- Real-time detection, distance estimation and spoken alerts
- Street field test in Bristol
- 2026Now
Foundation & validation
- Complete hazard models and accessible interface
- User pilot with an initial group
- Partnerships with sight-loss organisations
- 2027Next
Launch & growth
- Public beta for real-world feedback
- Institutional pilots with universities and rehab services
- Commercial launch
From the founder
“Technology should serve the people who need it most.”
Through market research, accessibility reports and conversations with potential users, I found the same gap again and again: people want to know what is directly in their path while they keep walking — without buying expensive hardware.
Sudha Rani MaddalaFounder, Vision Assist · MSc Data Science, University of Bristol
- Top 100Google Solution Challenge 2024 (team)
- Bristol PLUS AwardUniversity of Bristol
Questions
Frequently asked questions.
What is Vision Assist?
A smartphone app being developed to help blind and partially sighted people notice obstacles while walking. It uses the phone camera and AI to spot people, vehicles and obstacles, and tells you about them with short spoken alerts.
Does it replace my white cane or guide dog?
No. Vision Assist is designed to work alongside your usual mobility aid. It can warn about things a cane may miss, such as chest-height signs, but you should always keep using your cane or guide dog.
Can I download it today?
Not yet. Vision Assist is a working prototype. We are preparing a small pilot with blind and partially sighted users. You can join the pilot list on the Join page, and try the AI detection demo on this website now.
What phones will it work on?
The goal is standard modern iPhone and Android phones, with no extra hardware. Supported models will be confirmed before the pilot.
Will it cost money?
Core hazard alerts are planned to be free. Optional premium features and licences for organisations will help fund development, and we will look for sponsorship so price is never the barrier.
Help us build it right.
We’re looking for blind and partially sighted people to try Vision Assist in a small, supported pilot — and organisations who want to help.
Sign-ups go by email to the founder. Nothing is stored on this website.