ConductVision Enterprise
Weapon Detection
Identify firearms, edged weapons, and threat objects, with transparent accuracy reporting. Available for pilot.
Weapon Detection identifies firearms, knives, and other threat objects in video feeds to support security operations. The system is designed with full transparency about its capabilities and limitations: detection rates vary by weapon type, concealment level, and environmental conditions. It is intended as a supplementary layer for trained security personnel, not a replacement. All detection thresholds are configurable, and false positive rates are reported openly so operators can tune sensitivity to their specific risk tolerance. The system is built for environments subject to security screening requirements.
How it works
Weapon Detection in three steps
Connect Security Cameras
Position cameras at entry points, hallways, and open areas where weapons might be visible. The system works with existing surveillance infrastructure, no specialized cameras required.
Threat Object Classification
Detection models identify weapon-like objects and classify them by type (handgun, rifle, knife, blunt weapon). Confidence scores reflect the model's certainty, and detections below a configurable threshold are suppressed.
Alert Security Personnel
Detections above the confidence threshold trigger alerts to security operations centers with a snapshot, camera location, and classification. Security personnel make all response decisions, the system never initiates autonomous actions.
Features
Firearm Detection
Identifies handguns, rifles, and shotguns when visibly carried or brandished. Detection accuracy is highest for weapons held in open view and decreases for partially concealed items.
Edged Weapon Detection
Detects knives, machetes, and other bladed weapons. Performance is best on larger blades; small folding knives in hand are more challenging and reported with lower confidence.
Configurable Sensitivity Thresholds
Operators set detection confidence thresholds per camera and weapon type. Higher thresholds reduce false positives at the cost of some missed detections, and vice versa.
Transparent False Positive Reporting
The system logs every detection with its confidence score and provides daily, weekly, and monthly false positive rate reports. This data helps operators tune thresholds over time.
Human-in-the-Loop Design
All detections are routed to human operators for verification. The system explicitly does not make access control decisions or initiate lockdowns autonomously.
Integration with VMS Platforms
Outputs integrate with major video management systems (Milestone, Genetec, Avigilon) to highlight detection events in existing security operator workflows.
Use cases
School Security Augmentation
Schools need additional security layers beyond access control. Weapon Detection can provide an alert to security staff when a firearm is visibly present, supplementing existing security measures.
educationgovernment
Corporate Campus Security
Large campuses with multiple entry points cannot station security personnel everywhere. Camera-based detection can extend the coverage area that a security team effectively monitors.
corporate
Event Venue Screening Support
Concert venues and stadiums use metal detectors at entry. Camera-based detection can add a visual screening layer at perimeters and common areas beyond the entry checkpoint.
eventshospitality
Public Transit Security
Transit agencies monitor thousands of cameras across stations and vehicles. Automated detection can highlight feeds that require immediate security attention, reducing operator fatigue.
transportationsmart city
Retail Loss Prevention
Retail locations in high-risk areas benefit from an additional alert layer. Detection of weapons during robbery attempts can trigger silent alerts to law enforcement.
retail
Technical overview
- Models
- YOLOv8-M with weapon-specific training on vetted security datasets
- Input formats
- RTSP, ONVIF, MP4, HLS
- Output formats
- JSON, MQTT, Webhooks, VMS integrations (Milestone, Genetec, Avigilon)
- Edge support
- NVIDIA Jetson, Intel NUC, on-premise GPU servers
Accuracy and latency are measured against your footage and cameras during pilot scoping; we publish figures only with the conditions they were measured under.
Where this capability is used
Related capabilities
People Detection
Occupancy intelligence from the cameras you already have. Available for pilot at any scale.
Behavioral Analytics
Loitering, aggression, crowd surges, and abandoned objects, anomaly detection grounded in behavioral science. Available for pilot.
PPE Detection
Verify hard hats, vests, goggles, gloves, and respiratory protection, automatically, continuously. Available for pilot.
Evaluate Weapon Detection on your footage
One prioritized outcome, your cameras, a structured results review.
