On-device weapon detection.
Unprecedented at this price point.
YOLOv11 fine-tuned on legally-sourced security footage. Three new detection classes — handgun, rifle, knife — integrated directly into the RMADOR threat scoring pipeline. A weapon detection triggers a +80 threat score spike, immediate CRITICAL alert, and automatic 911 dispatch via RapidSOS — all within 3 seconds of the weapon appearing on camera.
How it works
Weapon detection is an extension of the base RF-DETR model, fine-tuned using Elastic Weight Consolidation (EWC, lambda = 5000) to prevent catastrophic forgetting of the original 7 detection classes. The model runs locally on the RMADOR server GPU — no footage is transmitted to any cloud service.
Weapon appears in frame
YOLOv11 detects the object class (handgun / rifle / knife) at the standard confidence threshold of 0.60. Detection is per-frame, per-camera, at 60+ fps.
Threat score spikes to CRITICAL
A weapon class detection contributes +80 to the Bayesian threat score immediately. Combined with zone multipliers and time-of-day weighting, scores reach 90–100 within a single assessment cycle.
Automated escalation
CRITICAL alert fires simultaneously: push notification to mobile (with vibration pattern), PagerDuty SEV-1, Twilio SMS/WhatsApp, and RapidSOS 911 dispatch — if score ≥ 90 and confidence ≥ 0.90 and enabled by Security Admin.
Detection classes
| Class | Threat Contribution | Automated Response |
|---|---|---|
| Handgun | +80 threat score | CRITICAL alert + 911 dispatch |
| Rifle | +80 threat score | CRITICAL alert + 911 dispatch |
| Knife | +80 threat score | CRITICAL alert + 911 dispatch (jurisdiction-dependent) |
Gunshot events (audio sensor) bypass normal threshold checks and immediately trigger CRITICAL regardless of threat score — weapon class detection follows the same bypass logic.
The economics comparison
Acoustic gunshot detection (ShotSpotter)
$65K – $500K
per city, per year
- ✕Audio-only — cannot identify weapon type or carrier
- ✕Detects after discharge — no pre-event warning
- ✕Requires city-scale infrastructure investment
RMADOR weapon detection (visual AI)
Included
in Enterprise tier
- ✓Detects weapon before discharge
- ✓Identifies class: handgun, rifle, or knife
- ✓Works on existing ONVIF cameras — no new hardware
- ✓Combines with audio sensor for redundant detection
ShotSpotter pricing based on publicly reported municipal contracts. RMADOR weapon detection is a software feature — no per-site infrastructure cost.
Use cases
Schools & Universities
Detect weapons at entry points before an incident escalates. Integrated with bollard lockdown and emergency door control.
Courthouses & Government Buildings
Supplement existing screening infrastructure with continuous AI coverage of all entry and exterior zones.
Sports Venues & Stadiums
Monitor crowd access points at scale. Weapon detection runs on existing CCTV — no new hardware required.
Critical Infrastructure
Power stations, water treatment facilities, and data centres where perimeter breach with a weapon demands immediate escalation.
Jurisdiction requirements
Weapon detection is a feature-gated capability that RMADOR enables per tenant. Deployment requires a Privacy Impact Assessment and legal review for your jurisdiction. The feature is disabled by default and cannot be self-activated.
- Canada (CA): Handgun and rifle detection permitted; knife detection requires additional legal review under PIPEDA.
- United States (US): All three classes available; state laws vary — legal review required per deployment.
- EU / UK / Australia: Requires national law compliance (GDPR Art. 9 / UK IPA / Australian Privacy Act) and explicit authority approval before activation.
RMADOR does not activate weapon detection without a signed legal review confirmation from the customer. Contact our team to begin the approval process.
Ready to add weapon detection?
Weapon detection is included in the Enterprise tier. Our team will guide you through the jurisdiction review and activation process.