Step 01Week 1
Mount & connect
A compact edge unit is placed at the display, shelf, or installation. It needs only power and a network uplink for events — no video cabling, no recording infrastructure.
Customer Experience
CrowdCanvas
Your display knows what it's showing. Now it can know if anyone cares. Anonymous audience measurement — presence, dwell, attention — processed entirely on the edge.
BuiltforSignage & DOOH operators, retailers, venues, museums, and artists running public-facing installations.
crowdcanvas.dunavnet.eu — Screen 12 · Atrium
How it works
From foot traffic to attention metrics in four steps — no faces, no footage.
CrowdCanvas turns a compact edge unit near your display or installation into an anonymous audience sensor. Video is processed in memory and destroyed within milliseconds; only counts and events leave the device.
Step 02Week 1
Define the scene
You define what matters: the passing zone, the viewing zone, the engagement zone. Choose the metrics — passers-by, stoppers, viewers, dwell time, group size.
Step 03Ongoing
Detect on the edge
Vision models run locally, detecting people as anonymous silhouettes. No faces, no demographics of individuals, no images stored or transmitted — audience insight without audience data.
Step 04Ongoing
Measure, compare, react
Attention metrics stream to your dashboard: passers vs. viewers, dwell distributions, conversion by daypart and content. Feed your CMS for proof-of-attention reporting — or drive the content itself in real time.
From frame to attention metric — and never a frame beyond the edge
Solution portfolio
See exactly what your team sees.
One sensor for signage, retail, and installations — the attention dashboard, creative comparisons and zone heatmaps your marketing and operations teams work with.
crowdcanvas.dunavnet.eu — Attention dashboard · Screen network
Integrates with your stack
Anonymous events, standard interfaces.
Feed signage CMS platforms for proof-of-attention reporting, BI tools for retail analytics, or creative engines for reactive installations.
Signage CMS
REST API
Webhooks
MQTT
Privacy by architecture. Video cannot leave the edge device — there is no pipeline for it to leave on. Only anonymous counts and events reach your CMS, BI, or creative engine. No faces, no demographics, no re-identification; GDPR & EU AI Act aligned, with full documentation for your DPIA.
Reference architecture
Your venue
Your environment
Display, shelf, venue or installation
Optional RTSP · local only
CrowdCanvas
CrowdCanvas edge
On-device inference · frames never stored
Anonymous events only
CrowdCanvas
CrowdCanvas cloud (optional)
Dashboards · reports · APIs
REST · WebSocket · OSC
Your venue
Your existing systems
Signage CMS · BI · creative engines
From the field
Where CrowdCanvas pays attention.
Real screens, real spaces — the edge unit at a display, a measured retail zone, an installation reacting to its audience.
Deploy three ways
Pick the model that matches your network and data policy.
SaaS
Hosted by us · EU region
Edge units on-site, dashboard in our EU cloud. Fastest start. Monthly billing per scene. ISO 27001-controlled.
- Live in 1–3 weeks
- Monthly · per scene
- EU data residency
Managed AzureMost chosen
Your tenant · we operate
Event platform inside your Azure tenant. Your data residency, your compliance posture, our operations team.
- Your tenant
- We run the ops
- Co-managed SLA
Custom / on-prem
Fully local
Everything on your network, including dashboards — suits venues and networks with strict data policies. Via the DunavNET Innovation Studio engagement model.
- Fully local option
- Custom integrations
- Studio engagement
Customer story
The metric the medium was missing.
European signage network
"For the first time we could tell an advertiser not how many times their spot played, but how many people actually stopped and watched it — and for how long. It changed how we price the network."
Network director
DOOH network · placeholder profile
48%Uplift in measured attention after creative optimisation
120kAnonymous audience events per site per month
2 wkFrom install to first report
0Frames of video stored or transmitted
Common questions
What operators and venues ask before they sign.
CrowdCanvas does not record, store, or transmit video, and it cannot identify anyone — it converts frames into anonymous counts on the device and discards them. This places it outside typical CCTV and biometric-analytics obligations. We provide full technical documentation to support your DPIA and legal review.
Not of individuals. Detections can be classified into broad age and gender classes on the device so you can see the composition of an audience in aggregate — no faces, no identity, no re-identification, and nothing that can be traced back to a person. The values are model detections and are meant as directional guidance for content planning.
Yes. Presence and attention events stream out over WebSocket, OSC, MQTT or webhooks, so creative engines like TouchDesigner, Unity, Unreal or Max/MSP can react to the audience as it arrives — not just report on it afterwards.
No. A compact edge unit is mounted at the display or installation and needs only power and a network uplink. Where a suitable camera already exists, it can be used over RTSP — but that stream stays local to your network.
Zone-level counts, dwell and attention rates are the metrics CrowdCanvas is tuned for, and they are validated on site during the pilot against your own observations. Figures are model detections — reliable for comparing screens, dayparts and creatives, not a person-by-person census.
SaaS is billed monthly per scene; Managed Azure and on-prem deployments are scoped per network. We'll size the right model and give you rough pricing in one 30-minute call.
Related solutions
