Frequently Asked Questions

CamGuard AI applies artificial intelligence to the security cameras your business already runs. It detects fire, smoke, people, and vehicles in the camera image. When a fire starts or someone enters without authorization, it alerts you within seconds by automated phone call, SMS, email, mobile push, and a dedicated on-site alarm device. All analysis happens on a GPU server inside your facility.

Yes. It works with any IP camera that supports ONVIF or RTSP. Analog cameras connect through an NVR or DVR that exposes an RTSP stream.

Brand doesn't matter: Hikvision, Dahua, Bosch, Axis, and equivalents all work. If you have no cameras or want to upgrade, we supply those too.

Analysis runs on a GPU server installed inside your facility. The camera stream never leaves the site.

Only confirmed alarms send an evidence frame and event record to the central panel. There is no continuous video upload, which keeps your bandwidth free and your data on-premises.

The system is built to process personal data in place. Footage is analyzed inside your facility, and only limited alarm evidence leaves the site. The panel provides per-user permissions, access logging, and retention controls.

Keep in mind that lawful video monitoring also carries operator-side duties, such as privacy notices and camera signage. We provide guidance during setup.

Once flame or smoke enters the camera's view, detection and verification usually finish within seconds and the alerts go out. Every camera is analyzed continuously, and each detection passes a second verification stage before it alarms.

Yes. The fire model recognizes both. In smoldering fires, smoke shows up long before flame does, which makes it the more valuable early signal.

A smoke detector waits for smoke to rise and physically reach the sensor. In open yards, high-ceiling warehouses, and greenhouses, that means it responds far too late or never triggers at all.

CamGuard AI watches the image instead. Fire and smoke are detected the moment the camera sees them, and you get evidence showing which camera and which spot. See the fire detection page for details.

A detection never becomes an alarm on its own. In sequence:

• Physical filters for object size, aspect ratio, and texture
• Static light filter that learns and removes lamps, LEDs, and reflectors
• Persistence and growth checks, since real fire flickers and spreads
• Day and night modes with separate thresholds for black-and-white footage
• Visual verification that confirms the event a second time

On top of that, per-camera zones and schedules exclude known sources like welding bays, stoves, and exhausts during specific hours.

It works at night, and thermal cameras are not required. Standard IP cameras with infrared night vision are enough.

The system detects when a camera switches to black-and-white night mode and applies night-specific brightness, texture, and shape criteria. That's why streetlights, reflectors, and headlights don't read as fire.

For each camera you define which zone is monitored and during which hours. An alarm fires when a person is detected outside permitted hours or inside a restricted zone.

This isn't a motion sensor. The system checks whether there's actually a person in frame, so swaying branches, animals, rain, and shadows don't trigger it. Vehicle detection can be set up as a separate rule. See the intrusion detection page for details.

Five channels fire at the same time:

• Automated phone call stating the event type and location
• SMS with a short summary
• Email with a snapshot, location, type, and timestamp
• Mobile push to panel users
• On-site alarm device with siren, flashing light, and LCD display

You configure which camera notifies whom, and when, from the panel.

It sounds a loud siren, flashes the fire or intruder icon in red, and displays the type and location of the incident on its LCD screen. Your staff knows within seconds, without looking at a phone.

Analysis runs on the on-site server, so cameras keep being monitored and the local alarm device still works. Outbound notifications like calls, SMS, and email are delivered once connectivity returns. No event records are lost.

A single on-site server analyzes dozens of cameras at once, depending on its GPU capacity, and hardware scales with your camera count. Businesses with multiple facilities install one server per site and manage them all from one central panel.

If your cameras are ready, setup usually finishes within one business day. You'll need:

• Space and power for a server on the same network as the cameras
• RTSP credentials for the cameras
• An internet connection so notifications can leave the site

We then register the cameras, draw the monitored zones, and set the time rules with you.

Yes. CamGuard AI doesn't replace your recording system. It adds an analysis layer on top. Your NVR keeps recording while CamGuard AI reads the same streams, analyzes them, and triggers alerts when something happens.

Yes. One server goes in per facility, and all sites report to a single central panel. Permissions work at the organization and user level, so a site manager sees only their own facility while headquarters sees everything.

Pricing depends on camera count, number of sites, the detection types you need, and hardware. Request a demo for a firm quote. We assess your site and give you a price built around it.

Yes. Fill out the demo form, email support@camguard.ai, or reach us at +90 555 595 26 00 by phone or WhatsApp. We demo the system live and design the right setup for your site together.