HawQ-ai · Predictive validation for AI perception systems

Right cameras. Right places. Right AI model. Proven.

One objective platform for every stakeholder: how many cameras, where, which optics, which AI — validated for coverage, blind spots, redundancy and AI behaviour on a digital twin of your site.

Book a demo Explore the products

Built in France · Inria spin-off

SENTINEL — SCENARIO CONSOLE SIM · 4 FEEDS

Analytics output DETECTED Nominal conditions — person detected.

Conceptual illustration — no measured data.

The product family

One platform, three products.

Design with Sentinel. Audit with Inspector. Harden with Synthetic.

AI video surveillance

Available

HawQ-ai Sentinel

Design new systems on a digital twin of the site — compare camera positions, lighting, RGB sensors and AI models; measure coverage, blind spots and AI performance.

See it in action ↓

HawQ-ai Inspector

Audit systems already deployed — real operational performance, low-performing zones, and the impact of the physical configuration on your vision models.

HawQ-ai Synthetic

Controlled attack and edge-case scenarios — rare, degraded, adversarial — to train, test and harden AI perception systems.

Adversarial patches
Glare & light
Fog & night
Tampering
Rare events

HawQ-ai Sentinel

From floor plan to powered-on system.

A 2D architectural plan transformed into a navigable 3D model of the building — A few photos → photorealistic twinConcept visual

Digital twin

A photorealistic twin of your site, from a few photos
Or start faster: a navigable 3D model from a 2D plan
On PC, no 3D expertise needed — the foundation every other step builds on

Camera configuration panel in the twin: sensor format, lens, field of view and resolution for real camera models — Sensor, lens, FOV & resolution — set in the twinPrototype output

Configure cameras

Pick real camera models — dome, turret, bullet
Set sensor format, lens, field of view and resolution
Swap hardware in one click, before buying it

Coverage dashboard over a reconstructed courtyard: covered and uncovered zones, camera distribution and key metrics — Coverage, blind spots & redundancyPrototype data

Validate performance

Coverage, blind spots and redundancy — zone by zone
Compare architectures: camera positions, models, mounting
Automatic positioning proposals, tuned to your purpose

Camera view with a person detected while an adversarial attack is flagged by the threat assessment panel — Detection under real conditions & attacksConcept visual

Validate the AI

Vary positions, orientations, optics and lighting
See where the AI holds and where it fails, zone by zone
Then run the attacks: glare, night, adversarial clothing, tampering

Installer wearing an XR headset in front of the installed camera, with AR panels showing camera info, cable route, field-of-view cone and coverage analysis — Installer view — XR guidance on siteConcept visual

Install in XR

Position, orientation and cable routes per camera
Expected POV and coverage, live in the installer's headset
Right the first time — fewer site revisits

Virtual expected POV compared with the actual VMS feed of the same camera, with a similarity score panel — Expected POV vs live VMS feed — similarity scoredConcept visual

Audit vs reality

Connect your VMS once the system is live
Compare the virtual footage with the real footage, camera by camera
Evidence for hand-over and compliance

01 / 05

Deliverables

Failure map
AI-drafted site study
XR install pack
As-built audit

Your footage never leaves your site.

Designed for sensitive and critical environments.

On-premise deployment
Air-gapped option
No telemetry
Video & models stored locally

See Sentinel on your own site.

Start from a floor plan, a few photos or a walkthrough video. Sentinel builds the twin, compares your options, and its AI agent drafts the study — objective, evidence-based, ready to sign. In the cloud or fully on-premise.

Book a demo