> For the complete documentation index, see [llms.txt](https://docs.imerit.net/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.imerit.net/3d-multi-sensor-fusion/labeling/3d-multi-sensor-fusion-labeling-editor/key-features/sanity-checks/cuboid-velocity.md).

# Cuboid Velocity

Cuboid Velocity is a sanity check that monitors how quickly a tracked cuboid moves between consecutive frames. When its velocity exceeds the configured threshold, the 3D Multi-Sensor Fusion Labeling Editor surfaces a visual cue so the annotation can be reviewed.

{% hint style="success" icon="lightbulb-on" %}
Velocity is evaluated using the selected axes. This lets you check movement in one direction, across a plane, or in all three dimensions.
{% endhint %}

***

## Configuring the Check

Open the project's category schema and add *Velocity* under the cuboid class's *Sanity Checks*. Configure the following values:

* **Axis** - choose `Vx`, `Vy`, or `Vz` for one axis; `Vxy`, `Vyz`, or `Vxz` for a plane; or `Vxyz` for three-dimensional velocity.
* **Velocity threshold** - enter the maximum permitted velocity. The value must be zero or greater.
* **Unit** - choose metres per second (`m/s`) or inches per second (`inch/s`).
* **Priority** - choose *High* for an error or *Low* for a warning.

Save the category schema after configuring the check. Changes apply to active tasks when the task is reloaded.

***

## Responding to a Velocity Violation

When a tracked cuboid exceeds the threshold on the selected axes, inspect its position in the affected frames. Correct a misplaced cuboid or adjust the configured threshold when the movement is valid for the dataset.

Velocity checks are non-blocking. Annotators can submit tasks with active warnings, and reviewers can use the displayed violations to identify annotations that need further inspection.

***

## Best Practices

* Select only the axes that matter for the scene. For example, use `Vxy` when vertical movement should not affect the check.
* Set thresholds using the dataset's frame rate and realistic speeds for each object class.
* Use the same unit as the project's measurement guidelines.
* Start with *Low* priority while calibrating thresholds, then use *High* for violations that should receive immediate attention.