---
title: "DetectionModel"
sidebar_label: "DetectionModel"
description: "The clika_runtime.modelverse.models DetectionModel class."
---

<!-- Generated by tools/api_reference/generate_api_docs.py. Do not edit. -->

A loaded object-detection model.

## `labels` (property)



## `open_vocabulary` (property)



## `video` (property)



## `detect`

```python
detect(self, image: 'clika_runtime.Tensor', *, confidence: 'float' = 0.5, max_detections: 'int' = 100, prompt: 'str | Any | None' = None, text_threshold: 'float | None' = None) -> 'Any'
```

The boxes in an [H, W, 3] image above ``confidence``. An
open-vocabulary model takes ``prompt`` (a text, or a prompt encoded
once with ``encode_prompt``).

## `detect_window`

```python
detect_window(self, frames: 'clika_runtime.Tensor', *, confidence: 'float' = 0.5, max_detections: 'int' = 100) -> 'list[Any]'
```

Detect over a window of frames ([F, H, W, 3]).

## `encode_prompt`

```python
encode_prompt(self, prompt: 'str') -> 'Any'
```

Encode a text prompt once, for reuse across images.