Task Settings

Task settings allow you to modify advanced features and adjust pre-configured settings during task creation, enabling you to do further calibrations to ensure you catch every occurring event.

Navigating to the Settings

After selecting a task under the Task Dashboard, you will be directed to that individual task's interface. Click Settings in the top right corner to enter the System Settings interface and adjust settings for that task.

Task Information

This tab contains metadata related to each task. These include:

• Task Name: The name of the task. Often this is set to be the camera's name in its video management system (VMS).

• Location: The location where the user's camera is set up. Read more on location setup here.

• Camera:

  • The camera stream used for the task, following an rtsp protocol. Read more on camera setup or upload multiple cameras at once here.

• Schedule: Either 24/7, daily, or weekly. Read more on configuring the task schedule here.

• Task Status: Indicates whether the dask is operational or disconnected, followed by the corresponding speed of the connection.

• Register: Contains the name of the user who created the task.

• Last Update: Indicates the date and time the latest modification to the task was made.

Camera View

By clicking Camera View on the left, you'l be able to access features such as previewing the live stream, adjusting the regions of interest for AI detection, and more adjustments for individual events.

A user is able to see the camera's real-time stream playback through the camera view. Press the blue play button in the middle of the camera view to start playing the stream.

Above the stream is an Edit button that the user can click into to reveal the following options:

• ROI (Region of Interest): The single focal area of the AI detection. As long as the bounding boxes of the desired ontology overlaps even partially within this ROI, it will be included in the detection. A possible area to omit from the ROI is any portion of the stream that does not include people when doing people-related counting tasks such as crowd density checking and trespassing detection.

  • To adjust the ROI, select ROI in the dropdown. When unmodified, the user should see a yellow dashed-line rectangle spanning the entire stream. Click the X symbol on the upper left of this polygon to delete and redraw the ROI. The drawing mechanism involves selecting the corners of the desired ROI polygon one by one. Make sure to double click on the last corner of the ROI polygon once satisfied. Finally, click Close Edit ROI to exit the ROI range editing mode. The final product will now show the areas excluded from the ROI to be darkened. The system will then apply the new range as the AI detection area.

• Excluded Zone: Allows the user to draw multiple polygons to mark areas excluded from the ROI. Compared to the ROI where the bounding box just needs to overlap even partially, the Excluded Zone requires that the center coordinates of the bounding box is within the Excluded Zone in order to be excluded from counting or detection.

  • To add or adjust Excluded Zones, select Excluded Zone in the dropdown. When adding excluded zones for the first time, the user can now directly draw on the stream by selecting the corners of the desired polygonal shape for the exclusion. Make sure to double click on the last corner of the polygon once satisfied. After drawing one excluded zone, the user can choose to draw more. Otherwise, click Close Edit Zone to exit the editing mode.

• Specific Event Area (e.g., Virtual Fences, Counting Area, etc.): When creating events that involve the use of a zone-and-entity overlap logic, a specific event zone gets created. One or more event areas may be present in the dropdown depending on how many events with different areas the user has created.

  • To modify this, select the zone in the dropdown. In this case, select Counting Area. An outlined and colored-in zone will appear. Click the X symbol on the upper left of this polygon to delete and redraw the zone. The drawing mechanism involves selecting the corners of the desired area/polygon one by one. Make sure to double click on the last corner of the polygon once satisfied. Depending on the settings you configured, this zone may be a single polygon or multi-polygon (as in, the user is only permitted to draw one or more polygons, like for the ROI and Excluded Zone respectively). When done, click Close Edit Zone to exit the editing mode.

Streaming Status

The streaming status allows the user to view stream health and any related information, such as:

• RTSP Source: Indicates the resolution and frame rate of the RTSP, which is recommended to be 720p and 10fps for optimal resource consumption and performance.

• Connection: Describes the connection between the camera and the detection server.

• Detection: The actual number of frames per second being used for AI detection.

Hint: The recommended RTSP resolution and frame rate are based off the fact that the AI model only processes 720p at 10fps. Anything above that simply gets downsized, which just adds an extra step and latency.

Image Collection

This feature allows the user to automatically save images based on specified AI object threshold values, preserving valuable resources for future training.

View the current conditions for collecting images under the Image Collection tab, including:

• Snapshot Interval: Defines how long to wait before saving the next image. Storing images intensively can accumulate a large number of similar photos, which does not help in obtaining diverse data for AI training. It is recommended to save at least once every 5 minutes.

• Confidence Threshold: Set this field to save images when the AI-detected object's confidence score falls within this range. The system can set the storage range for different object categories.

• Max Collection: The maximum number of images to collect from a task. For cameras running 24/7, they may store a large number of similar images. Setting a maximum collection number can reduce the number of similar pictures captured from the same camera task.

• Image Tag: Sets a tag to your images to help you quickly find them when searching for AI training data images in the future.

After modification, click Save to apply the new settings.

Video Recording

Enabling the video recording features allows the user to view short clips of the event occuring. Note that the video recording option is only available for object detection events, or object detection + VLM events.

Video recording is turned off for all events by default. To enable the feature, select the Edit button on the lower right quadrant in the image above to toggle the video recording switch for all the events.

Once turned on, the next events detected with the video recording feature turned on will look similar to the case below. After the AI and image view, a video recording view is now available.

As opposed to the view below without the video option.

Notification

The Notification tab allows the user to edit the event notification recipients.

Once saved, notifications will be received via email and also in the Observ interface.

Email notifications include the task involved, the timestamp, the objects that the AI detected, the image of the event, and a direct link to the task.

Observ notifications can be accessed on the upper right quadrant of the interface. Note that notifications do not appear here until manually activated for each task.

Clicking on a notification opens a pop-up window of the event record where the user can see the event name, description, camera, location, and time of occurrence. You can toggle the display of bounding boxes for each detected object on the upper right corner.

Scrolling down the event window reveals the Status dropdown menu, which you can use to report the event as correct or a false alarm. The system will record this for future reference and data analysis.