Frequently asked questions.

Occivar is a privacy-first edge AI platform that anonymously counts people, measures density and surfaces anomalies in physical spaces, without ever storing or transmitting video. The platform runs on small edge devices co-located with cameras, producing anonymous occupancy counts that feed dashboards, BMS, Slack and reports.

Buildings, campuses and venues need accurate, real-time occupancy data to right-size HVAC and lighting, prove sustainability targets, manage crowd safety and design space, but traditional CCTV creates privacy and compliance risk, and badge swipes only measure entry, not actual occupancy. Occivar gives operators the occupancy number they need without ever creating an identity-bearing record.

Traditional CCTV analytics typically uploads video frames to a cloud server for processing, which creates privacy, bandwidth and compliance exposure. Occivar runs the people-counting model on-device; raw frames never leave the camera, only anonymous counts and density metrics are emitted. There is no video stored, no biometric data, no demographic inference.

Property and facilities teams, sustainability leads, security and operations managers, council and university planners, transport operators and event venues, anyone who needs to know how a space is used in real time without taking on the legal or ethical risk of identifying the people in it.

Occivar is an Australian deep-tech company headquartered in Melbourne, Victoria. The company was founded in 2025 and operates under ABN 29 691 189 134.

No. Occivar does not use facial recognition, biometric identification or demographic inference under any circumstance. The model detects the presence of a person as an anonymous count; no face embeddings, gait signatures or other identifiers are ever generated.

No. Raw video frames are processed on the edge device and discarded in memory; they are never written to disk, never uploaded to the cloud, and never sent anywhere. Only the resulting anonymous occupancy counts and density values leave the device.

Occivar is architected to avoid collecting personal information altogether; the platform produces anonymous occupancy counts only. Because no personal information is collected, the Australian Privacy Principles obligations that apply to identifiable data do not trigger. We can share an architecture letter that procurement and legal teams can include in their assessment.

Occivar produces only anonymous aggregate counts, so the platform falls outside the scope of GDPR Article 4(1) personal-data definitions in most deployments. For deployments inside the EU we can supply a technical architecture letter that procurement and DPO teams can attach to their record-of-processing assessment.

Privacy at Occivar is enforced by architecture, not policy. The on-device inference pipeline can be inspected, the network egress is anonymous counts only, and the platform never receives the data it claims not to have. We publish the architecture detail on the Privacy by design page and welcome procurement audits.

In most deployments yes, the edge device is co-located with the existing camera and the camera continues to operate as it normally does. Where the existing camera also feeds a security DVR or NVR, Occivar operates independently and does not require access to the recorded footage.

Edge AI runs the machine-learning model on a device near the camera, instead of streaming video to a server. Occivar uses edge AI because it is the only architecture that can produce a real-time occupancy count while honestly claiming that no video left the device. Read the deep dive at /technology/edge-ai.

The edge inference runtime targets common embedded boards in the Jetson, Raspberry Pi 5 and Hailo accelerator families. We size the device to the venue: a single Raspberry Pi can serve several low-traffic zones, while busy concourses use a Jetson Orin class device.

Counting accuracy depends on camera angle, lighting and the calibration of the zone definition; in typical indoor deployments we target ±5% steady-state and ±10% during dense peaks, against a ground-truth manual count. We do not over-claim accuracy; see the engineering post on counting accuracy and ground truth.

Yes, outdoor and low-light deployments are supported with appropriate camera selection. Outdoor zones need cameras with adequate dynamic range and weather rating; the inference pipeline itself is unchanged.

Talk to your data is a conversational interface on top of the occupancy data. Operations leads can ask plain-English questions like "how was the library used last Friday" and get charts, schedule recommendations or exportable reports back. The assistant only sees the same anonymous counts the dashboard sees. See the live demo at /technology/talk-to-your-data.

A single-floor pilot is typically commissioned in two to four weeks: one week to scope the zones and confirm camera coverage, one week on-site for device install and calibration, and a one-to-two week shake-out before reports go live. Larger multi-zone rollouts are scoped per site.

Yes, anonymous occupancy counts can drive HVAC setpoints, lighting schedules and ventilation rates via standard BMS protocols (BACnet, Modbus) or via a REST integration layer on top. See the Integrations page for the full list of supported endpoints.

Occivar routes anonymous counts and threshold alerts into Slack, Microsoft Teams, Snowflake, S3, email and the major BMS controllers. The Alerts layer also supports custom webhooks for niche destinations.

Often no, existing cameras at the correct angle and resolution can be used. Where a zone needs additional coverage, we will scope a small number of additional cameras as part of the deployment. The platform does not require a uniform camera fleet.

The Occivar dashboard is a web application served from a privacy-first hosting region (currently Sydney for Australian customers). Authentication is per-tenant; data from one customer is never visible to another.

Occivar is deployed across universities, councils, retail, health and aged care, events and stadiums, transport, corporate offices and construction sites. The same privacy-first architecture is tuned to the operational questions each kind of space asks. See the full industries directory.

Yes, anonymous occupancy data per zone over a multi-week trailing window lets a property team show, hour by hour, which opening windows are running empty and which are over-subscribed. The HVAC and lighting savings model is independent of the pilot data. Read the methodology at right-sizing library hours.

Yes, the platform measures density per zone, tracks egress under defined thresholds, and integrates threshold alerts into the operations channel during an event. See the events and stadiums page for the queue, arena and exit/evacuation flows.

Yes, the platform gives a corporate property team anonymous, per-zone, per-hour occupancy data that a badge log cannot produce. Badges measure entry; occupancy measures presence. Read the field-notes post on office utilisation without a badge log.

Yes, anonymous occupancy monitoring is well suited to safety-by-default deployments in aged-care common areas and corridors, where biometric or video-based monitoring would be unacceptable. See /industries/health-and-aged-care.

The fastest path is a single-floor or single-zone pilot: we scope the zones with your facilities team, install one edge device per zone, calibrate against a manual ground-truth count, and run a four-week trailing report. Book a demo to start the conversation.

Pricing depends on the number of zones, the camera mix and the integration scope. We work to a transparent per-zone licence model with a one-off commissioning fee. Smaller pilots are sized to land inside a discretionary procurement threshold so the proof-of-value conversation can happen quickly.

Our case studies page tracks the live pilots and wins. The team is currently scoping its first paying pilots in Australia; we are forward-looking about pilot status and do not over-claim results we have not yet delivered.

For pilot enquiries, partnerships or technical questions, use the contact page or request a demo. We respond from Melbourne business hours.