Obseron AI plugin installation and usage

Obseron AI plugin enables object recognition using CPU or GPU. Object recognition can be used to trigger recording or rules.

Obseron AI plugin requires an additional AI component license, to purchase a license please contact your closest distributor.

Obseron AI plugin only works in Windows environment.

Obseron AI plugin installation

• 1. Download AI-plugin GPU- tai CPU-package from Obseron download page https://www.obseron.com/download (GPU -version requires an Nvidia GPU with CUDA support)
  2. From Obseron, open Settings > Plugins, click the + button on the bottom of the window, select the downloaded AI plugin package and press Open.
  3. Once the AI plugin is done installing, you can find its settings from the Plugins page. Use GPU can be toggled if the installed package was the GPU package and if the server is equipped with Nvidia GPU. By default the detection rate is 1000ms, meaning the detections will happen once every second. Faster detection rate increases the accuracy but also the CPU/GPU load.

Adding and using Obseron AI component

1. 1. 1. Open Settings > Cameras and select the camera you want to add the AI component to
      2. Open the selected cameras Components tab and add an Obseron AI component
      3. Once the component is added, the settings for this component will be shown
         
         • Object detection tab
           • Select the objects to detect, person and/or car.
           • Select the confidence threshold, this threshold defines how confident the recognition must be for it to be accepted. Received events list above shows each confidence level with every event, this value can be used to tweak the conficence level and to avoid false detections.
           • Select the detection point, by default Bottom center. The detection point defines the object trigger location on the area.
         • Presence tab
           • Optionally create an area for the Presence condition and set how long the object is required to be on the area for the condition to trigger.
         • Entering tab
           • Optionally create an area for the Entering condition, condition triggers when the recognized object enters the area.
         • Exiting tab
           • Optionally create an area for the Exiting condition, condition triggers when the recognized object exits the area.
      4. Now a rule can be created to use these conditions
         
         • Create a new rule
           • Select Camera event condition, select the used camera, set event type to Presence.
           • Select any action, for example Notification/Alarm.
      5. Areas and detection squares can be shown on the camera image by selecting Areas from the radial menu while in Edit mode.
      6. There is also an AI events tab in the log, where you can see the confidence percentages of each recognition with thumbnails.