# AI-R: Model Switching & Confidence Control

AI-R now gives you greater control over your detection models enabling operators to configure detection accuracy and performance to their environment in real time.

This release introduces two powerful new controls:

1. **Model Selection:** Switch between *Larger Object Detection* and *Smaller Object Detection*
2. **Confidence Threshold Adjustment:** Fine-tune your detection sensitivity directly from the dashboard letting you decide if AI-R prioritizes catching every possible detection or filtering for only high-confidence results.

### Configuring Detection Behaviour

In the **AI-R Settings** panel inside the Cockpit View, you’ll now find:

#### Model Type Dropdown

* **Large Object Detection Model** – optimized for close-range, high-confidence detection of people and vehicles.
* **Small Object Detection Model** – optimized for identifying people and vehicles at longer ranges or when targets appear smaller in wide-area surveillance missions.

<figure><img src="https://1057699820-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FCGWcCPxNn1SifRiAjnAW%2Fuploads%2FPCUzqHvQj6tNtwTHeYHM%2FScreenshot%202025-10-09%20at%208.29.03%E2%80%AFPM.png?alt=media&#x26;token=747161b7-d95a-4ae8-8d3b-504a22de2d8d" alt=""><figcaption><p>AI Detection Settings</p></figcaption></figure>

<figure><img src="https://1057699820-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FCGWcCPxNn1SifRiAjnAW%2Fuploads%2F7fE2KwiWXphJuZufwNo4%2FModel_Selection.png?alt=media&#x26;token=b7b652df-798b-44f0-b230-0f7574ddec39" alt=""><figcaption><p>Detection model selection</p></figcaption></figure>

### Confidence Threshold Slider

The **Confidence Threshold** slider lets you choose a value between **0% and 100%**, determining how confident the AI must be before labeling an object as detected.

* Lower values = more detections (less strict).
  * Useful when you don’t want to miss anything Eg, detecting small or far-away objects.
* Higher values = fewer detections (higher precision).
  * Useful for precision-critical environments. Eg, perimeter security or automated alert

<figure><img src="https://1057699820-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FCGWcCPxNn1SifRiAjnAW%2Fuploads%2FPTUAepu69h6lcfDSlxfi%2FScreenshot%202025-10-09%20at%206.36.10%E2%80%AFPM.png?alt=media&#x26;token=d5b131d8-ca2e-4f66-8ecb-fa879fc81abd" alt=""><figcaption><p>Adjusting Confidence Threshold</p></figcaption></figure>

Please feel free to reach out to **<support@flytbase.com>** for more information on FlytBase AI-R.
