Opening
Imagine navigating this space without being able to see it.
The interactive presentation opens inside a busy mainline train station at rush hour. An optional soundscape plays the station as it is heard, not seen. It contains:
- Rolling suitcase wheels over stone flooring
- Dozens of overlapping conversations, none of them directed at the traveller
- Fast footsteps passing close on both sides
- A distant, half-audible platform announcement
- Train doors opening, then a chime and closing
- Traffic and rain from the street entrance behind
Every day millions of people navigate spaces like this effortlessly. For over 250 million people living with severe vision loss, it can be overwhelming.
The problem
Vision loss is not an edge case. It is a daily navigation problem at scale.
- 55,000 people with severe vision loss in Denmark — our first market, and the community we build with directly.
- Over 250 million people worldwide living with moderate to severe vision impairment or blindness.
Existing solutions
Every existing tool solves part of the problem.
The white cane
Reliable, trusted, and unchanged for seventy years.
Strengths
- Always works
- No battery, no updates
- Universally understood
Where it falls short
- Only detects what is within one metre of the ground
- No warning for overhead or approaching hazards
- Cannot tell you what something is
Phone apps
Powerful recognition, awkward in motion.
Strengths
- Rich object and text recognition
- Low cost
- Familiar hardware
Where it falls short
- Occupies a hand that is already holding a cane
- Requires aiming the camera manually
- Depends on a network connection
Today's AI glasses
Impressive demos, built for sighted users.
Strengths
- Hands-free capture
- Conversational answers
- Socially discreet
Where it falls short
- Describe scenes, but do not understand space
- Send video to the cloud to think
- Not designed around orientation and mobility
None of them do all of it. That gap is where Visionar sits.
The product
Two parts. A head unit that senses, and a pack that thinks.
A lightweight wearable in two parts, developed so that depth sensing, computer vision and conversational AI can run entirely on the device.
The concept design shows a matte black glasses frame with an integrated bridge camera and two computer-vision cameras at the outer corners, a speaker built into the temple arm, and a single cable running to a small rectangular compute pack that clips onto clothing, worn off the face.
- Part one — the head unit. A glasses-mounted unit carrying the cameras, microphone and speaker: everything that has to see and hear the room, worn at eye and ear level.
- Part two — the compute pack. A small clip-on pack running the AI entirely on-device, so the system works without an internet connection and keeps personal data on the device.
In use
What it does for the person wearing it.
A short demonstration video shows a walkthrough of the glasses in an everyday environment. The system itself is being built to do five things:
1. Spatial awareness — knowing what is ahead, and how far away it is
The glasses build an understanding of the surroundings and can register obstacles the wearer would otherwise meet without warning — including things at head height that a cane never touches.
2. Works anywhere — no signal required
The system runs on the device itself, so it keeps working in stairwells, underground stations and anywhere else a connection drops out.
3. Private by design — nothing leaves the device
Because the glasses do not send anything to a server, what the wearer and the people around them do stays with them. Privacy comes from how the product is built, not from a policy.
4. Ask and be answered — a spoken question, a useful answer
Wearers can ask about what is in front of them and hear a short answer through a discreet speaker that leaves normal hearing free.
5. Hands stay free — alongside the cane, not instead of it
Nothing has to be held or pointed. Existing mobility aids stay in use, and guidance arrives only when there is something worth saying.
Competitive landscape
Capable devices exist. Spatial understanding that runs offline does not.
Existing products excel at reading text or describing a scene. Few combine three-dimensional perception with processing that stays on the device.
| Product | Origin | Stage | Sensing | Processing | Approximate price |
|---|---|---|---|---|---|
| Visionar (ours) | Denmark | Prototype in development | 3D depth and vision | Fully on device | DKK 12,000 to 17,000 |
| OrCam MyEye | Israel | Commercial | 2D only | Cloud dependent | About DKK 24,000–37,000 |
| Envision Glasses | Netherlands | Commercial | 2D only | Cloud dependent | About DKK 25,000 |
| Meta Ray-Ban | USA | Mass-market consumer | 2D only | Cloud dependent | About DKK 1,500–2,000 |
| .lumen (dotLumen) | Romania | Pre-commercial pilots | 3D | Mostly on device | About DKK 75,000 |
| NOA (biped.ai) | Switzerland | Commercial | 3D | Cloud dependent | About DKK 18,000 |
| WeWALK Smart Cane | UK and Turkey | Commercial | Cane based | App based | Lower cost |
Market
One country first. Then one certification.
The assistive-technology market for vision loss is generally reported to grow at roughly 11–13% a year, driven by ageing populations, better wearable AI, and the European Accessibility Act in force since 2025.
- Denmark — approximately 55,000 people with significant vision loss, of whom approximately 20,000 are immediately addressable: working-age, mobile adults. Primary market, with statutory device funding via Serviceloven §112.
- United Kingdom — approximately 2 million people. Active field-testing partnership with the Middlesex Association for the Blind.
- Europe (EU and EEA) — approximately 30 million people, reachable through one certification process.
- Global — approximately 250 million people. Long-term context only.
More than 2.2 billion people live with some form of vision impairment worldwide, according to the WHO.
Revenue model
One near-term line through public funding. Three that add upside later.
Near-term income depends only on device sales through an already established public funding route. The other three lines are not required for this stage to succeed.
Line one — device sales (near term)
The device is sold to the individual user and paid through the public reimbursement pathway rather than out of pocket. In Denmark this is Serviceloven §112, under which municipalities fund approved assistive devices for citizens with a permanent disability. A retail channel through an optician chain is a possible second route.
Line two — research and academic hardware access (potentially earlier)
Hardware made available to university and industry labs working on spatial AI or wearable sensing — similar in spirit to Meta's Project Aria research kit, where approved partners receive hardware in exchange for research collaboration. It does not depend on a large installed base, and would give structured usability feedback alongside the MAB field testing.
Line three — venue mapping services (follows adoption)
The device works best in buildings that have already been mapped. Visionar is developing a self-service tool that lets venue staff map and update their own building in ten to twenty minutes per session. Venues pay for access.
Line four — spatial data licensing (long term)
Over time, maps from the venue mapping tool would form a dataset with value to companies building delivery robots and warehouse automation, which currently rely on expensive manual surveying. Realistic only once enough venues have been mapped.
Route to adoption
Denmark already has a funded pathway from prototype to practical use.
Approximately 55,000 people in Denmark live with significant visual impairment, and the country's reimbursement system and accessibility ecosystem provide an established route to reach them.
- Step one — Serviceloven §112. Danish municipalities may provide approved assistive devices to citizens with a permanent disability, so the device is publicly funded rather than paid out of pocket.
- Step two — HMI Basen. Once regulatory requirements are met, inclusion in the national database of approved assistive technologies gives municipalities a clear route to procure and distribute the device.
- Step three — certification by design. CE marking, product classification, data protection and accessibility standards guide technical decisions during prototype development rather than after it.
- Step four — collaborative validation. Universities, municipalities, rehabilitation specialists, accessibility organisations and future users are involved in validating the technology in practice.
| Stakeholder | Role | Timing |
|---|---|---|
| MAB / Dansk Blindesamfund | User feedback and testing | Immediate |
| Municipalities | Reimbursement and procurement under §112 | Mid-term, post-certification |
| Independent visually impaired adults | End users | Mid-term |
| Public institutions | Pilot mapping partners | Near-term |
| Retail chains | Adoption requires a validated inclusivity or revenue case | Long-term |
| Robotics and spatial-data companies | Potential licensees of aggregated, de-identified mapping data | Long-term, exploratory |
Roadmap
A staged path that reduces technical uncertainty before each next step.
- Initial physical prototype — completed. Core AI software developed and validated on commercially available hardware, including an Intel RealSense depth camera.
- Custom hardware prototype — in progress. A wearable prototype integrating custom-selected sensors and embedded AI hardware for real-time on-device processing. This is the current project.
- User testing and iteration — future. Evaluation with visually impaired users covering navigation performance, usability and comfort, followed by iterative refinement.
- Certification and pilot preparation — future. Regulatory preparation, manufacturing considerations, and pilots with accessibility organisations and municipalities.
- Deployment and market introduction — long term. Initial deployment in Denmark, followed by expansion to additional European markets.
Eight months of hardware development
Our development plan for the next 8 months, month by month:
| Months | Milestone | Deliverable |
|---|---|---|
| 1–2 | Component procurement | All candidate camera modules, depth sensors, compute boards, audio and power components ordered and received. |
| 3–4 | Baseline benchmarking | Each candidate component tested against wearable power, weight, heat and latency constraints; comparative results documented. |
| 5 | Final configuration and rig assembly | Winning components selected; first custom sensor rig assembled in a 3D-printed housing. |
| 6 | Data collection and model training | Training data collected on the custom rig; on-device AI models adapted to the new sensor configuration. |
| 7 | Prototype refinement | Housing, mounts and power management iterated; refined wearable prototype produced. |
| 8 | Field-testing preparation | Prototype and test protocol finalised and handed over for field testing with the Middlesex Association for the Blind. |
Funding
Each grant has funded one clearly defined stage.
Skylab, supported by the Bevica Foundation — DKK 20,000 (received)
Funded the initial physical prototype on commercially available hardware.
Internationalization Fund — up to DKK 30,000 (received)
Supports participation in Innovation Camp Shanghai in January 2027 for market research and international networking.
Next funding — to be confirmed
Further funding will be sought for user testing, hardware refinement, certification and pilot collaborations.
Team
Built at DTU, CBS and RUC — with the people we build for.
- Noah Savgu — CEO, hardware and partnerships — noah@visionar.dk
- Akira Adeniran-Lowe — CTO, computer vision and architecture — akira@visionar.dk
- Max Stalzer — data and MLOps, embedded AI — max@visionar.dk
- Ava Roshan — commercial strategy — ava@visionar.dk
- Mona Zainab Mustafa — business development and social impact — mona@visionar.dk
In collaboration with MAB, Bevica Fonden, KU Lighthouse, the Copenhagen School of Entrepreneurship and DTU Skylab.
Guidance comes from hardware founders and researchers in computer vision and embedded systems — Nikolaj Grathwol, Theodora Kontogianni and Charalampos Orfanidis — alongside MAB, DTU Skylab, KU Lighthouse and the Copenhagen School of Entrepreneurship. The defining principle stays the same: technology developed with people who are vision-impaired, not only for them.
Vision
We believe independence begins with knowing what is around you.
Visionar — AI that helps people navigate the world with confidence.
Contact
Thank you.
Building the future of independent navigation.
Looking for: deep-tech and assistive-tech investors, pilot partners in reimbursement-funded healthcare systems, and technical peers in embedded AI.
- Email: info@visionar.dk
- Web: visionar.dk
- Based in Copenhagen, Denmark