BOX 3.2: creating a site and camera and configuring a preset

Create a site and virtual camera, select a preset and check that events appear.

Updated 27 September 2026 · Application administrator · 13 min

1. Prepare the condition and video recording

This example looks for a person without a helmet. The condition must be observable: the person's head is distinguishable and the required protective helmet is absent. An obscured head is a reason to review the source video, not confidently record a missing helmet.

For repeatable training, use a virtual camera with a clip. This lets you review the same episode without waiting for a real incident. For an IP camera, select a supported source type and its connection parameters instead of a file.

BOX 3.2: the Sigma camera view separates Camera and Neural network output and provides tabs for zones, archive and performance. Check both video acquisition and analysis output. Localized reference view; captured settings and values are preserved.

2. Create a site

Open “Settings → Sites”. Create a site with a clear name in the correct part of the tree. Save it and find it in the list again. A site groups sources by location; the camera name alone does not create this assignment.

Check that the correct parent site is selected. For a large organization, it is useful to reflect the physical structure: organization, workshop, area. Use clear names that the operator will recognize in the log.

BOX 3.2: the new-site form, with name, parent path, map and image fields. This blank form explains the inputs; it does not show a saved Sigma site.

3. Add a virtual camera

Open camera creation under the selected site. On the “General information” tab, select “Virtual camera”, enter the camera name and choose a video file. The training example used the available helmet demonstration MP4 clip. For your own video, upload it and wait for the file to appear in the selection.

Check the site in the “Location” section. Open the video: its content must match the task, and the image must not be an interface capture or title screen. Check the beginning, relevant episode and end of the recording.

The camera and neural network have separate switches. Analysis requires both the source and neural network to be enabled. Looping playback can be selected for a repeatable exercise. It may generate new events on every loop, which is expected for a training source.

Click “Save”. If the form does not close or the operation responds slowly, find the camera in the tree first. Do not create a second camera just because the interface did not return to the list.

BOX 3.2: an unsaved Virtual camera form with camera name, File selection and upload controls. No file is selected and no camera has been created in this view.

BOX 3.2: the lower part of the unsaved camera form selects the site and workspace and exposes separate Camera state and Neural network controls. Both are OFF; this does not show an active or tested camera.

4. Configure detectors and a preset

On the “Detectors” tab, select a suitable preset. For example, use “Helmet monitoring” to detect missing helmets. Check the selected categories and parameters, then save the camera.

A preset stores a set of analytics settings. The decision matrix determines how results are processed and classified. These are related but different settings: assigning a preset does not replace checking how a detection becomes a notification.

Do not transfer numeric thresholds from another scene without testing. A person's height in pixels, lighting and occlusion have a strong effect on quality. Examine the source frame and detections first, then change parameters one at a time.

BOX 3.2: a new preset with its name, category search and parameter columns. The displayed switches are off; this is a field reference, not the configured Sigma preset.

BOX 3.2: the Detectors tab before a preset has been selected. Add a preset, then verify its enabled categories; the empty form does not demonstrate active detection.

5. Restrict the monitoring area

If monitoring is required only in a specific area, configure a zone on the camera image and its purpose. The region must match the actual work area; a neighboring walkway must not become part of the zone unnoticed.

Recheck the annotations after changing resolution, camera angle or position. A polygon on screen does not prove that the selected detector uses it: test the rule on episodes inside and outside the zone.

Sigma's first training event was obtained outside the marked zones. This confirms the basic source-to-analytics chain; zones are checked in a separate exercise.

BOX 3.2: the camera Zones tab before any zone has been added. Use Add zone to open the editor, then select the zone type required by the task. Localized reference view; captured settings and values are preserved.

6. Confirm that events arrive

Open “Logs → Violations”. Select the training camera and a period covering its operation. Find a notification by camera name, time and type. Check the image: it must come from the selected clip.

The training run produced “Helmet (violation)” entries. Their initial status was “Not reviewed”. This is a detector result for the operator to review.

The “Sigma — production area” training camera in the violation log. Localized reference view. Original camera frames and event states are unchanged. Shown with administrator access; available controls depend on the assigned role.

If the log is empty

Check the chain in order. The same empty log may mean no violations, an incorrect filter or disabled processing.

Check	Expected state
Source	The video opens and contains the required episode
Switches	The camera and neural network are enabled
Settings	The preset and categories match the task
Zone	The object falls within the monitored area under the rule used
Log	Correct camera, period and statuses
Permissions	The user has access to the source and log

BOX 3.2: Camera performance shows separate timelines for video acquisition and inference. The grey Disabled intervals in this example are not evidence of healthy continuous processing. Localized reference view; captured settings and values are preserved.

Use the documentation and configuration supplied for your installed release. Environment addresses and access credentials are provided privately with the delivery.