# Track bounding boxes

## Overview

A common issue when doing object detection on video is that detections are not associated between each camera frame. This can cause issues for use cases such as counting items in a video. In essence without tracking, each frame would would add the the count.

## Implementation

We have implemented a version of ByteTrack to track object detection outputs between frames.

## Usage

The most common usage is to directly connect the output of an object detection component to this component.

<figure><img src="https://3816131924-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F1SuYrwusqikNVYCnh3Rw%2Fuploads%2F9fUQq19dEQI3Z9u6JOwi%2FScreenshot%202023-11-05%20at%2012.40.38%E2%80%AFPM.png?alt=media&#x26;token=39eac1c4-8417-4268-95e4-cfe70d4b67c5" alt=""><figcaption></figcaption></figure>

## Using the output

The component modifies each bounding box that is passed into it and adds/updates the following fields on the bounding box.&#x20;

&#x20;

#### Bounding box reference

{% content-ref url="../../data-types/bounding-box" %}
[bounding-box](https://docs.cloneable.ai/cloneable-documentation/platform-documentation/data-types/bounding-box)
{% endcontent-ref %}