UrbanVue
UrbanVue is a Dutch computer-vision platform that inspects public space from ordinary street imagery, scoring litter, road damage, greenery and street furniture against CROW quality levels A to D with GPS-located, automatically anonymised evidence.
What is UrbanVue?
UrbanVue is a public-space inspection platform built by UrbanVue B.V., a Rotterdam start-up whose tagline, AI kijkt, jij beslist, means that the AI looks and the manager decides. Instead of sending inspectors to walk a sample of predefined locations, it runs computer vision over thousands of ordinary street images and grades what it finds against the CROW image quality standards of the 2023 Kwaliteitscatalogus Openbare Ruimte, the same A (good) to D (poor) scale Dutch municipalities already use in their maintenance contracts. The homepage counters claim more than 200 detection classes across more than 32 image measures. Three product lines share one platform. Waste covers litter and bulky refuse, bags dumped beside containers, graffiti and stickers announced at up to 95% accuracy, special waste such as nitrous oxide canisters and abandoned white goods, blocked gullies, and before-and-after cleanliness measurement. Roads covers cracks, potholes and alligator cracking, loose or subsided paving, traffic signs, markings, leaning poles, manhole covers and gullies, all read from ordinary dashcam or GoPro footage without LiDAR and without disrupting traffic. Greenery covers weeds along edges and around obstacles, tree pits, root growth, leaf fall, grass height and overgrowth. A fourth strand covers street furniture: lamp posts, RVV traffic signs, utility cabinets, hydrants, charging points and bins, scored on condition and inventoried with GPS for asset management. The chain runs from the UrbanVue Camera app through automatic anonymisation into the platform. People, vehicles and number plates are pixelated by a panoptic segmentation model before analysis, and raw images are deleted within twenty-four hours. What remains is an interactive map coloured by quality level, filters on capture date, confidence, category and level, time series per district, frame-by-frame routes back to the source image, open exports, role-based access per organisation, and a two-click hand-off into Signalen, the reporting system many Dutch municipalities already run. Named references include the municipalities of Rotterdam, Delft, Wassenaar and Eemsdelta, contractor Krinkels and The Hague University of Applied Sciences.
What it does
- Detect litter, dumped waste, bulky refuse, graffiti and stickers on ordinary street imagery
- Score every finding against the CROW image quality levels A to D and place it on the map by GPS
- Inspect road surfaces for cracks, potholes, loose paving, subsidence, manhole covers, signs and markings
- Assess street greenery: weeds, tree pits, root growth, grass height, overgrowth and leaf fall
- Spot leaning lamp posts and poles and inventory street furniture with its exact position
- Blur people, vehicles and number plates automatically before any analysis is run
- Turn a detection into a Signalen report in two clicks, or export the whole set to GIS
When to use UrbanVue / When not to
A quick filter to help you decide if UrbanVue is the right fit.
When to use UrbanVue
- Dutch municipal public-space managers who already grade their streets against the CROW image quality standards
- Outdoor maintenance teams that want to target cleaning and repair rounds instead of running fixed routes
- Policy makers and budget holders who need the quality spread across a district expressed in figures rather than impressions
- Maintenance contractors and asset managers checking condition against an agreed quality level
- Organisations with no inspection hardware budget, since an ordinary smartphone is enough to collect the imagery
When not to use UrbanVue
- Teams outside the Netherlands: the A-to-D scale, the PAS 2161 work and the Signalen link are all Dutch references
- Anyone who needs an English interface, as the website, the platform and both apps are Dutch only
- Developers expecting a documented public API, since exchange runs through file exports
- Buyers who want a published price or a self-service sign-up before speaking to anyone
- Anyone looking for surveillance or law-enforcement imagery, a use the vendor rules out explicitly
How to use UrbanVue
A typical end-to-end flow, from setup to results.
- Book an introductory call and agree which areas and which CROW measures matter for your management question
- Receive access to the UrbanVue Camera app and go through the privacy points together
- Install the app on an ordinary smartphone and sign in through Keycloak single sign-on
- Capture street imagery on foot, by bike or from a vehicle, with GPS and timestamp written into the EXIF automatically
- For vehicle runs, switch to Camera Pro on Android with interval capture, manual exposure, a speed gate and cockpit mode
- Let the upload queue drain: it works offline, can be limited to WiFi and can delete photos once they are sent
- As an alternative, hand over existing image libraries or GoPro footage the municipality already holds
- Wait for processing, which normally puts results in the platform within the hour after capture
- Open the map, filter by category, date, confidence and quality level, and walk each route frame by frame
- Confirm, correct or reject detections, then export them to your GIS or push them into Signalen
Pros & Cons
Pros
- Scoring anchored in a public, recognised standard, so results compare directly with a manual inspection
- Fine-grained coverage, since every image in an area is analysed rather than a handful of sample points
- No hardware to buy: an ordinary smartphone replaces LiDAR or a dedicated survey vehicle
- Privacy handled at the source, with anonymisation before analysis and a DPIA rated positively by a municipality
- Hosting and model training on the vendor's own servers in the Netherlands, with no third-party AI service in the image chain
- Open exports plus a direct link into Signalen, so a finding reaches the team that has to act on it
- Human control is structural: the platform signals and scores but schedules nothing and sends nothing on its own
Cons
- No published pricing and no terms and conditions, so both budgeting and contracting start with a sales conversation
- Dutch only, across the website, the platform and both mobile apps
- The value depends on Dutch frameworks: CROW levels, PAS 2161 and Signalen have no direct equivalent abroad
- No documented public API, so interoperability runs through file exports rather than a live connection
- ISO 27001 is cited as a framework and an alignment, not as a certification the company holds
- Models are improved on anonymised customer material and no opt-out is documented
- Thin published evidence for a young company: one accuracy figure, and two different areas quoted for the same one-hour turnaround
Pricing & Plans
UrbanVue does not publish prices. There is no pricing page, no plan list and no amount anywhere on the site; the commercial route is a demonstration request or the contact form, with a reply promised within one working day. Neither a permanent free plan of the service nor a free trial is advertised. The only free component is the capture app: UrbanVue Camera is listed at EUR 0 on both the Apple App Store and Google Play. No price is stated for UrbanVue Camera Pro.
- No named pricing plan is published on the site
- free capture app for iOS and Android
- professional Android capture app for in-vehicle collection
- price not stated
- Waste
- Roads and Greenery are functional product lines on a single platform
- not pricing tiers
- A pilot on a defined area is the entry point the vendor proposes
Data, GDPR & hosting
A consolidated view of how UrbanVue handles your data.
GDPR overview
Implementation is documented in unusual detail for a company of this size. The privacy statement, version 2.0 of March 2026, runs to thirteen sections and names a legal basis for every operation: consent, performance of a contract, legitimate interest and legal obligation. A data protection impact assessment has been carried out and was assessed positively by the municipality of Delft. Data subject rights are listed with their articles, from access under Article 15 to withdrawal of consent under Article 7(3), with requests handled by e-mail under the subject line AVG-verzoek and answered within thirty days, and a route to the Dutch supervisory authority. All servers sit inside the European Economic Area. Under the AI Act the vendor places itself in the minimal-risk category. No data protection officer is named, and no Article 27 representative is designated since the company is established in the European Union.
Who owns the data?
Under the GDPR, UrbanVue splits its role in two. For street imagery it acts as a processor: the commissioning municipality is the controller, owns the material and gives the instructions, and an Article 28 processing agreement is in force for each municipality. For its own website, platform accounts and app accounts, UrbanVue is the controller. Detection results, meaning location, type and severity, are shared with that municipality; the vendor states it sells no personal data and shares none with advertisers. Anonymised material may be reused by UrbanVue to improve its own detection models. No terms and conditions document is published, so the privacy statement of March 2026 is the only public source on ownership.
Reuse rights
Detections, locations and severity scores belong to the commissioning municipality and can be taken out at will: the platform exports to JSON, Shapefile, CSV, GeoJSON, GeoPackage and Excel for any GIS or asset system, and pushes findings into Signalen as reports. Export files are deleted automatically after seven days. No licence condition on reuse is published, for the simple reason that no terms and conditions document exists on the site, so reuse arrangements are settled in the contract rather than online. On its own side UrbanVue reuses only anonymised material to improve its detection models, trains locally on its own servers inside the European Economic Area, sends nothing to external AI services, and states that no personal data is used for training. No opt-out from that model improvement is documented.
Data retention & training
Hosting summary
All infrastructure is managed by UrbanVue itself on its own servers in the Netherlands, and every server sits inside the European Economic Area. The database, application servers and storage services are not reachable from outside, and raw imagery is held in folders isolated per organisation. Three subprocessors are disclosed: Keycloak for identity management on the vendor's own server in the Netherlands, Complianz for cookie consent running locally, and YouTube via Google Ireland Limited for embedded video, loaded only after explicit consent. The only processing that may reach the United States is that YouTube path, covered by the EU-US Data Privacy Framework; no street imagery, platform data or app data leaves the EEA. Detection models are trained locally on those same servers, without any external AI service. Note that the public website itself is fronted by a Cloudflare anycast address, which geolocates outside the EEA; that is the content delivery layer of the marketing site, not the application hosting.
Things to keep in mind
Risks and trade-offs to weigh before adopting UrbanVue.
- Street imagery is personal data until it is blurred, and the duty to inform residents before a capture run sits with the municipality, not with UrbanVue
- The municipality is the controller for that imagery, so it carries the legal responsibility and must sign a processing agreement before anything is collected
- Anonymised customer material feeds model improvement and no opt-out is documented, which is worth settling in the contract
- A quality score can quietly become a target: chasing an A everywhere may pull crews away from problems the model does not detect
- Inspectors who stop walking the street lose the contextual judgement the vendor itself says the tool cannot replace
- Short retention cuts both ways, since anonymised frames disappear after thirty days and a disputed score cannot be re-examined afterwards
- ISO 27001 is a stated framework and full GDPR compliance is the vendor's own claim; neither has been verified by a third-party audit
Setup & Integrations
Technical difficulty
Low for the customer. The vendor promises an operational setup within a week, with no IT project and no special hardware: an ordinary smartphone is enough. In practice it means installing a mobile app, signing in through a Keycloak account that is provided, and collecting imagery. Nothing is installed on the customer's servers. The real effort sits elsewhere, in the organisational work: agreeing the privacy points, informing residents, choosing a pilot area, and configuring the Signalen link and the GIS exports with the teams that will use them.
Deployment
Apps stores
Integrations
Supported languages
Behind UrbanVue
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
What standard does UrbanVue score against?
How is this different from a traditional street inspection?
What imagery can it work with?
How is privacy handled?
How long is the imagery kept?
Where is the data hosted?
Does it connect to municipal reporting systems?
Is there an API, and what does it cost?
Should you pick UrbanVue?
UrbanVue is a narrow tool, and that is its strength. By grading street imagery against the CROW image measures rather than inventing a scale of its own, it turns an inspection that was a subjective sample into a measurement a municipality can defend in a budget discussion. The coverage argument is convincing: every image in an area is read, not a handful of points, and the hardware barrier is genuinely low, since a smartphone in a jacket pocket replaces a survey vehicle. Governance is handled better than the average of this market. Anonymisation happens before analysis rather than after, raw imagery is deleted within a day under audit logging, hosting and model training stay on the vendor's own servers inside the European Economic Area, no external AI service touches the images, and the AI Act position is documented with material supplied on request for an algorithm register. A DPIA assessed positively by the municipality of Delft is a stronger signal than most self-declarations. The reservations are commercial rather than technical. Nothing about the price is public, no terms and conditions document exists, there is no documented API, and everything, from the website to both apps, is in Dutch. The product is bound to Dutch frameworks, so its usefulness drops sharply outside the Netherlands. Published evidence is still thin: a single accuracy figure, and two different area figures for the same one-hour turnaround. This is a young company, with a domain registered in September 2024 and a first web archive capture in February 2025, though its named references are real and verifiable. For a Dutch municipality or contractor already working with image measures, it deserves a demonstration. For anyone else, it is worth watching rather than buying.
- Choosing a selection results in a full page refresh.
- Opens in a new window.