> For the complete documentation index, see [llms.txt](https://nx.docs.scailable.net/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://nx.docs.scailable.net/ai-manager-v6.1.4/ai-manager-plugin/configure-the-plugin/postprocessors/left-behind-object-detection.md).

# Left Behind Object Detection

Illegal dumping detection works by comparing the bounding boxes of a model to a reference. Detected objects which are not in the reference are tracked and timed. If these objects persist for longer than the threshold time they are flagged as dumped objects.

Left behind objects are presented as standard bounding boxes with the 'dumped' type. Create an event in Nx Meta to raise an alarm when a 'dumped' type object is detected.

By default, the AI Manager creates a reference image from the first frame it receives on startup. This reference image will then be saved and reused in future runs. It is also possible to manually create a reference image.

When the frame is empty of temporary objects and a good representation of the background is being displayed, use the **Trigger Reference Run** button to set the new reference. Once the button is pressed, the following frame(s) will be used to set a new reference.

<figure><img src="/files/D0tNhU70qacjUZBm0RzX" alt="Left Behind Object Detection postprocessor settings panel"><figcaption><p>Left Behind Object Detection settings</p></figcaption></figure>

**Dumping Detection Class:** The class of objects to run the postprocessor on. This is useful if you have a model which detects multiple types of objects, but you only want to check for left behind objects on a specific class, for example 'backpack'.

**Dumping Detection Time Threshold:** The amount of time ( in seconds ) an object of the desired class has to be in frame and completely stationary before it is flagged as a dumped object.

**Trigger Reference Run:** A button to trigger a reference run. When clicked, the following frames create a new reference and discard the previous one.

Left behind object detection is compatible with any model which generates bounding boxes.