### Install SAHI from Source

Source: https://obss.github.io/sahi/quick-start

Install SAHI directly from its GitHub repository. This is useful for getting the latest development version.

```bash
pip install git+https://github.com/obss/sahi.git@main
```

--------------------------------

### Install Ultralytics Framework

Source: https://obss.github.io/sahi/quick-start

Install the Ultralytics framework, a common choice for object detection models used with SAHI.

```bash
pip install ultralytics
```

--------------------------------

### Install Project Dependencies

Source: https://obss.github.io/sahi/contributing

Install core and development dependencies using `uv sync`. For specific model testing, additional dependencies may be required.

```bash
# Install core + dev dependencies
uv sync --extra dev

# For testing specific models, install their dependencies.
```

--------------------------------

### Clone SAHI Repository and Install Dependencies

Source: https://obss.github.io/sahi/notebooks

Clone the SAHI repository from GitHub and install the necessary development dependencies to run the notebooks locally. This setup is required before launching Jupyter.

```bash
git clone https://github.com/obss/sahi.git
cd sahi
pip install -e "[dev]"
jupyter notebook demo/
```

--------------------------------

### Install SAHI

Source: https://obss.github.io/sahi/quick-start

Install SAHI using pip. This is the primary method for adding SAHI to your project.

```bash
pip install sahi
```

--------------------------------

### Install YOLOv5 Package

Source: https://obss.github.io/sahi/guides/models

Install the yolov5 package to use classic YOLOv5 models.

```bash
pip install yolov5

```

--------------------------------

### Install MMDetection Packages

Source: https://obss.github.io/sahi/guides/models

Install mmdet, mmcv, and mmengine to use the full MMDetection model zoo.

```bash
pip install mmdet mmcv mmengine

```

--------------------------------

### Install HuggingFace Transformers and TIMM

Source: https://obss.github.io/sahi/guides/models

Install the transformers and timm packages to use object detection models from the HuggingFace Hub.

```bash
pip install transformers timm

```

--------------------------------

### Get Package Information

Source: https://obss.github.io/sahi/api

Retrieves whether a package is installed and its version string. Logs the version if verbose is True. Returns 'N/A' for the version if the package is not installed.

```python
import importlib.metadata as _importlib_metadata
import importlib
import logging
from packaging import version

logger = logging.getLogger(__name__)

def get_package_info(package_name: str, verbose: bool = True) -> tuple[bool, str]:
    """Check whether a package is installed and retrieve its version.

    Args:
        package_name: The name of the package to look up.
        verbose: If True, log the package version when available.

    Returns:
        is_available (bool): Whether the package is installed.
        version_string (str): Version string, or "N/A" if not installed.
    """
    _is_available = is_available(package_name)

    if _is_available:
        try:
            _version = _importlib_metadata.version(package_name)
        except (ModuleNotFoundError, AttributeError):
            try:
                _version = importlib.import_module(package_name).__version__
            except AttributeError:
                _version = "unknown"
        if verbose:
            logger.pkg_info(f"{package_name} version {_version} is available.")
    else:
        _version = "N/A"

    return _is_available, _version
```

--------------------------------

### Install RF-DETR

Source: https://obss.github.io/sahi/guides/models

Install the RF-DETR library using pip. This is a prerequisite for using RF-DETR models.

```bash
pip install rfdetr

```

--------------------------------

### Install PyTorch and TorchVision

Source: https://obss.github.io/sahi/guides/models

Install torch and torchvision to use built-in TorchVision detection models like Faster R-CNN and RetinaNet.

```bash
pip install torch torchvision

```

--------------------------------

### Install Detectron2 Package

Source: https://obss.github.io/sahi/guides/models

Install the detectron2 package to use Facebook's Detectron2 models for detection and instance segmentation.

```bash
pip install detectron2

```

--------------------------------

### Development Installation of SAHI

Source: https://obss.github.io/sahi/quick-start

Install SAHI in editable mode for development. This allows you to make changes to the SAHI code and see them reflected immediately without reinstalling.

```bash
git clone https://github.com/obss/sahi
cd sahi
pip install -e .
```

--------------------------------

### Install SAHI with Conda

Source: https://obss.github.io/sahi/quick-start

Install SAHI using Conda, suitable for environments where Conda is the preferred package manager.

```bash
conda install -c conda-forge sahi
```

--------------------------------

### Install Ultralytics for YOLO Models

Source: https://obss.github.io/sahi/guides/models

Install the ultralytics package to use YOLOv8, YOLO11, YOLO26, and other Ultralytics model variants.

```bash
pip install ultralytics

```

--------------------------------

### Display SAHI Environment Information

Source: https://obss.github.io/sahi/cli

Display installed package versions related to SAHI. This command provides a quick way to check your SAHI installation and its dependencies.

```bash
sahi env

```

--------------------------------

### Get MMDetection Version as Integer

Source: https://obss.github.io/sahi/api

Retrieves the installed MMDetection version and converts it into an integer format for easier comparison or processing.

```python
def mmdet_version_as_integer() -> int:
    """Get the MMDetection version as an integer."""
    import mmdet

    return int(mmdet.__version__.replace(".", ""))
```

--------------------------------

### __init__

Source: https://obss.github.io/sahi/api

Initializes the postprocessor with configurable matching parameters such as threshold, metric, and classagnostic behavior.

```APIDOC
## __init__ (match_threshold=0.5, match_metric='IOU', class_agnostic=True)

### Description
Initialize the postprocessor with matching configuration.

### Parameters
#### Path Parameters
- **match_threshold** (float) - Optional - Minimum overlap value to consider predictions matching. Default: 0.5
- **match_metric** (str) - Optional - Metric for overlap computation, "IOU" or "IOS". Default: 'IOU'
- **class_agnostic** (bool) - Optional - If True, apply postprocessing across all categories. Default: True
```

--------------------------------

### __init__

Source: https://obss.github.io/sahi/api

Initializes the Detectron2 detection model, setting up required packages.

```APIDOC
## __init__

### Description
Initializes the Detectron2 detection model, setting up required packages including torch and detectron2.

### Parameters
#### Path Parameters
None

#### Query Parameters
None

#### Request Body
None

### Parameters
- `*args` (object) - Variable length argument list passed to DetectionModel.
- `**kwargs` (object) - Arbitrary keyword arguments passed to DetectionModel.
```

--------------------------------

### Check Package Availability and Version

Source: https://obss.github.io/sahi/api

Use this function to determine if a Python package is installed and to get its version. It handles cases where the version might be unknown or outdated, raising an ImportError if a minimum version is not met.

```python
from sahi.utils.import_utils import get_package_info

_is_available, _version = get_package_info(package_name, verbose=verbose)
if _is_available:
    if _version == "unknown":
        logger.warning(
            f"Could not determine version of {package_name}. Assuming version {minimum_version} is compatible."
        )
    else:
        if version.parse(_version) < version.parse(minimum_version):
            raise ImportError(
                f"Please upgrade {package_name} to version {minimum_version} or higher to use this module."
            )
yield
```

--------------------------------

### Install PyTorch and Ultralytics for CUDA

Source: https://obss.github.io/sahi/quick-start

Install PyTorch, torchvision, pytorch-cuda, and ultralytics in a single Conda command for CUDA environments. Ensure compatibility by installing them together.

```bash
conda install -c pytorch -c nvidia -c conda-forge pytorch torchvision pytorch-cuda=11.8 ultralytics
```

--------------------------------

### __init__(*args, **kwargs)

Source: https://obss.github.io/sahi/models/detectron2

Initializes the Detectron2 detection model. It sets up the required packages and calls the parent class constructor.

```APIDOC
## __init__(*args, **kwargs)

### Description
Initialize Detectron2 detection model.

### Parameters
#### Path Parameters
- `*args` (object) - Required - Variable length argument list passed to DetectionModel.
- `**kwargs` (object) - Required - Arbitrary keyword arguments passed to DetectionModel.
```

--------------------------------

### __init__

Source: https://obss.github.io/sahi/models/yolov5

Initializes the YOLOv5 detection model. It sets up the required packages and calls the superclass constructor.

```APIDOC
## __init__

### Description
Initializes YOLOv5 detection model.

### Method Signature
`__init__(*args, **kwargs)`

### Parameters
* `*args`: Variable length argument list.
* `**kwargs`: Arbitrary keyword arguments.

### Source Code
```python
def __init__(self, *args: object, **kwargs: object) -> None:
    """Initialize YOLOv5 detection model."""
    existing_packages = getattr(self, "required_packages", None) or []
    self.required_packages = [*list(existing_packages), "yolov5", "torch"]
    super().__init__(*args, **kwargs)  # type: ignore[misc, arg-type]
```
```

--------------------------------

### Create and Activate Python Virtual Environment

Source: https://obss.github.io/sahi/contributing

Set up a Python 3.10 virtual environment using `uv` and activate it. This ensures project dependencies are isolated.

```bash
pip install uv
uv venv --python 3.10
source .venv/bin/activate  # On Windows: .venv\Scripts\activate
```

--------------------------------

### __init__

Source: https://obss.github.io/sahi/models/base

Initializes the object detection/instance segmentation model with various configuration options.

```APIDOC
## `__init__`

### Description
Init object detection/instance segmentation model.

### Parameters
#### Path Parameters
None

#### Query Parameters
None

#### Request Body
None

### Parameters
- **model_path** (str | None) - Optional - Path for the instance segmentation model weight
- **model** (Any | None) - Optional - A pre-loaded detection model instance.
- **config_path** (str | None) - Optional - Path for the mmdetection instance segmentation model config file
- **device** (str | None) - Optional - Torch device, "cpu", "mps", "cuda", "cuda:0", "cuda:1", etc.
- **mask_threshold** (float) - Optional - Value to threshold mask pixels, should be between 0 and 1. Default: 0.5
- **confidence_threshold** (float) - Optional - All predictions with score < confidence_threshold will be discarded. Default: 0.3
- **category_mapping** (dict | None) - Optional - Mapping from category id (str) to category name (str) e.g. {"1": "pedestrian"}
- **category_remapping** (dict | None) - Optional - Remap category ids based on category names, after performing inference e.g. {"car": 3}
- **load_at_init** (bool) - Optional - If True, automatically loads the model at initialization. Default: True
- **image_size** (int | None) - Optional - Inference input size.

### Request Example
```python
from sahi.models.base import BaseSahiModel

model = BaseSahiModel(model_path="path/to/your/model.pt", config_path="path/to/your/config.py")
```

### Response
#### Success Response (200)
None

#### Response Example
None
```

--------------------------------

### Initialize Legacy Postprocessor

Source: https://obss.github.io/sahi/api

Initializes the postprocessor with configuration for matching predictions. Supports 'IOU' or 'IOS' as matching metrics and allows for class-agnostic matching.

```python
def __init__(
    self,
    match_threshold: float = 0.5,
    match_metric: str = "IOU",
    class_agnostic: bool = True,
) -> None:
    """Initialize the postprocessor with matching configuration.

    Args:
        match_threshold: Minimum overlap value to consider predictions matching.
        match_metric: Metric for overlap computation, "IOU" or "IOS".
        class_agnostic: If True, apply postprocessing across all categories.
    """
    self.match_threshold = match_threshold
    self.class_agnostic = class_agnostic
    if match_metric == "IOU":
        self.calculate_match = self.calculate_bbox_iou
    elif match_metric == "IOS":
        self.calculate_match = self.calculate_bbox_ios
    else:
        raise ValueError(f"'match_metric' should be one of ['IOU', 'IOS'] but given as {match_metric}")
```

--------------------------------

### check_package_minimum_version

Source: https://obss.github.io/sahi/api

Checks if an installed Python package meets a specified minimum version requirement. It returns True if the package is missing, its version is unknown, or if the installed version meets or exceeds the minimum. It returns False if the installed version is below the minimum.

```APIDOC
## check_package_minimum_version(package_name, minimum_version, verbose=False)

### Description
Check whether an installed package meets a minimum version requirement.

### Parameters
#### Path Parameters
- **package_name** (str) - Required - The name of the package to check.
- **minimum_version** (str) - Required - The minimum acceptable version string (e.g. "1.0.0").
- **verbose** (bool) - Optional - If True, log the detected package version. Defaults to False.

### Returns
- **bool** - True if the package is missing, its version is unknown, or its version meets the minimum. False if the installed version is below the minimum.
```

--------------------------------

### Initialize Postprocessor

Source: https://obss.github.io/sahi/postprocess/backends

Initializes the postprocessor with configuration parameters for matching.

```python
def __init__(
    self,
    match_threshold: float = 0.5,
    match_metric: str = "IOU",
    class_agnostic: bool = True,
) -> None:
    """Initialize the postprocessor with configuration parameters.

    Args:
        match_threshold: Minimum overlap value (IoU or IoS) to consider
            two predictions as matching.
        match_metric: Overlap metric, "IOU" or "IOS".
        class_agnostic: If True, apply postprocessing across all
            categories. If False, apply per category independently.
    """
    self.match_threshold = match_threshold
    self.class_agnostic = class_agnostic
    self.match_metric = match_metric

```

--------------------------------

### __init__

Source: https://obss.github.io/sahi/api

Initialize MMDetection detection model.

```APIDOC
## __init__

### Description
Initialize MMDetection detection model.

### Parameters
- **model_path** (str | None) - Optional - Path to the model file.
- **model** (object | None) - Optional - Pre-loaded model object.
- **config_path** (str | None) - Optional - Path to the model configuration file.
- **device** (str | None) - Optional - Device to run the model on (e.g., 'cuda', 'cpu').
- **mask_threshold** (float) - Optional - Threshold for mask prediction. Defaults to 0.5.
- **confidence_threshold** (float) - Optional - Threshold for confidence score. Defaults to 0.3.
- **category_mapping** (dict | None) - Optional - Mapping for category IDs.
- **category_remapping** (dict | None) - Optional - Remapping for category names.
- **load_at_init** (bool) - Optional - Whether to load the model at initialization. Defaults to True.
- **image_size** (int | None) - Optional - Expected image size.
- **scope** (str) - Optional - Scope for the model. Defaults to 'mmdet'.
```

--------------------------------

### Install Dev Dependencies for CI Fix

Source: https://obss.github.io/sahi/contributing

Install development dependencies to fix CI build failures related to formatting.

```bash
uv sync --extra dev
```

--------------------------------

### Download YOLOv5-Nano Model

Source: https://obss.github.io/sahi/api

Downloads the YOLOv5-Nano model to a specified or default path. Creates parent directories if they do not exist.

```python
def download_yolov5n_model(destination_path: str | None = None) -> None:
    """Download the YOLOv5-Nano model for testing."""
    if destination_path is None:
        destination_path = Yolov5TestConstants.YOLOV5N_MODEL_PATH

    Path(destination_path).parent.mkdir(parents=True, exist_ok=True)

    if not path.exists(destination_path):
        urllib.request.urlretrieve(
            Yolov5TestConstants.YOLOV5N_MODEL_URL,
            destination_path,
        )

```

--------------------------------

### Initialize Yolov5DetectionModel

Source: https://obss.github.io/sahi/models/yolov5

Initializes the Yolov5DetectionModel, adding 'yolov5' and 'torch' to required packages.

```python
class Yolov5DetectionModel(DetectionModel):
    """YOLOv5 object detection model.

    Wraps Ultralytics YOLOv5 for fast object detection.
    """

    def __init__(self, *args: object, **kwargs: object) -> None:
        """Initialize YOLOv5 detection model."""
        existing_packages = getattr(self, "required_packages", None) or []
        self.required_packages = [*list(existing_packages), "yolov5", "torch"]
        super().__init__(*args, **kwargs)  # type: ignore[misc, arg-type]
```

--------------------------------

### Get and Set Category Name

Source: https://obss.github.io/sahi/utils/coco

Use the `category_name` property to get or set the name of the annotation's category. The setter ensures that the provided category name is a string.

```python
@property
def category_name(self) -> str | None:
        """Returns category name of the annotation as str."""
        return self._category_name

    @category_name.setter
def category_name(self, n: str) -> None:
        if not isinstance(n, str):
            raise Exception("category_name must be a string")
        self._category_name = n
```

--------------------------------

### sahi.cli.app

Source: https://obss.github.io/sahi/api

Command-line interface application entry point for SAHI.

```APIDOC
## `app()`

### Description
Entry point for the SAHI command-line interface application.

### Function Signature
```python
def app() -> None:
    """Cli app."""
    fire.Fire(sahi_app)
```
```

--------------------------------

### Get and Set Category ID

Source: https://obss.github.io/sahi/utils/coco

The `category_id` property allows you to get and set the category identifier for the annotation. The setter includes a type check to ensure that the `category_id` is always an integer.

```python
@property
def category_id(self) -> int:
        """Returns category id of the annotation as int."""
        return self._category_id

    @category_id.setter
def category_id(self, i: int) -> None:
        if not isinstance(i, int):
            raise Exception("category_id must be an integer")
        self._category_id = i
```

--------------------------------

### Check Package Minimum Version

Source: https://obss.github.io/sahi/api

Verifies if an installed package meets a specified minimum version. Logs the detected version if verbose is True. Returns False only if the installed version is below the minimum.

```python
def check_package_minimum_version(package_name: str, minimum_version: str, verbose: bool = False) -> bool:
    """Check whether an installed package meets a minimum version requirement.

    Args:
        package_name: The name of the package to check.
        minimum_version: The minimum acceptable version string (e.g. "1.0.0").
        verbose: If True, log the detected package version.

    Returns:
        True if the package is missing (assumed compatible), its version
            is unknown, or its version meets the minimum. False if the
            installed version is below the minimum.
    """
    from packaging import version

    _is_available, _version = get_package_info(package_name, verbose=verbose)
    if _is_available:
        if _version == "unknown":
            logger.warning(
                f"Could not determine version of {package_name}. Assuming version {minimum_version} is compatible."
            )
        else:
            if version.parse(_version) < version.parse(minimum_version):
                return False
    return True

```

--------------------------------

### Initialize SAHI CLI App

Source: https://obss.github.io/sahi/api

Initializes the command-line interface application for SAHI using the fire library. This function serves as the entry point for SAHI's CLI.

```python
def app() -> None:
    """Cli app."""
    fire.Fire(sahi_app)
```

--------------------------------

### Detectron2 Model Initialization

Source: https://obss.github.io/sahi/api

Initializes the Detectron2 detection model. This is the entry point for using Detectron2 models with Sahi.

```APIDOC
## `__init__(*args, **kwargs)`

### Description
Initialize Detectron2 detection model.

### Parameters
* `*args`: Positional arguments.
* `**kwargs`: Keyword arguments.
```

--------------------------------

### get_package_info

Source: https://obss.github.io/sahi/api

Checks if a specified Python package is installed and retrieves its version. It can optionally log the package version if verbose is set to True. Raises an ImportError if the installed version is less than a specified minimum version.

```APIDOC
## get_package_info(package_name, verbose=True)

### Description
Check whether a package is installed and retrieve its version.

### Parameters
#### Path Parameters
- **package_name** (str) - Required - The name of the package to look up.
- **verbose** (bool) - Optional - If True, log the package version when available. Defaults to True.

### Returns
- **is_available** (bool) - Whether the package is installed.
- **version_string** (str) - Version string, or "N/A" if not installed.
```

--------------------------------

### Initialize DetectionModel in SAHI

Source: https://obss.github.io/sahi/api

Initializes the DetectionModel with various configuration options. Use 'load_at_init=True' to automatically load the model upon instantiation. Ensure necessary dependencies are met.

```python
class DetectionModel:
    """Base class for all detection models in SAHI.

    Subclasses must implement ``load_model``, ``perform_inference``, and
    ``_create_object_prediction_list_from_original_predictions`` to integrate
    a new detection framework. The base class handles device management,
    dependency checking, category remapping, and the public prediction API.
    """

    required_packages: list[str] | None = None

    def __init__(
        self,
        model_path: str | None = None,
        model: Any | None = None,
        config_path: str | None = None,
        device: str | None = None,
        mask_threshold: float = 0.5,
        confidence_threshold: float = 0.3,
        category_mapping: dict | None = None,
        category_remapping: dict | None = None,
        load_at_init: bool = True,
        image_size: int | None = None,
    ) -> None:
        """Init object detection/instance segmentation model.

        Args:
            model_path: str
                Path for the instance segmentation model weight
            model: Any
                A pre-loaded detection model instance.
            config_path: str
                Path for the mmdetection instance segmentation model config file
            device: Torch device, "cpu", "mps", "cuda", "cuda:0", "cuda:1", etc.
            mask_threshold: float
                Value to threshold mask pixels, should be between 0 and 1
            confidence_threshold: float
                All predictions with score < confidence_threshold will be discarded
            category_mapping: dict: str to str
                Mapping from category id (str) to category name (str) e.g. {"1": "pedestrian"}
            category_remapping: dict: str to int
                Remap category ids based on category names, after performing inference e.g. {"car": 3}
            load_at_init: bool
                If True, automatically loads the model at initialization
            image_size: int
                Inference input size.
        """
        self.model_path = model_path
        self.config_path = config_path
        self.model: Any = None
        self.mask_threshold = mask_threshold
        self.confidence_threshold = confidence_threshold
        self.category_mapping = category_mapping
        self.category_remapping = category_remapping
        self.image_size = image_size
        self._original_predictions: Any = None
        self._object_prediction_list_per_image: list[list[ObjectPrediction]] | None = None
        self._batch_images: list[np.ndarray] | None = None
        self._original_shapes: list[tuple[int, ...]] | None = None
        self.set_device(device)

        # automatically ensure dependencies
        self.check_dependencies()

        # automatically load model if load_at_init is True
        if load_at_init:
            if model:
                self.set_model(model)
            else:
                self.load_model()

    def check_dependencies(self, packages: list[str] | None = None) -> None:
        """Ensures required dependencies are installed.

        If 'packages' is None, uses self.required_packages. Subclasses may still call with a custom list for dynamic
        needs.
        """

```

--------------------------------

### mmdet_version_as_integer

Source: https://obss.github.io/sahi/api

Get the MMDetection version as an integer.

```APIDOC
## mmdet_version_as_integer()

### Description
Get the MMDetection version as an integer.

### Returns
- **int** - The MMDetection version as an integer (e.g., 20201 for v2.0.2).
```

--------------------------------

### CocoAnnotation.iscrowd

Source: https://obss.github.io/sahi/api

Gets the iscrowd flag for the annotation.

```APIDOC
## CocoAnnotation.iscrowd

### Description
Returns iscrowd info of the annotation.

### Property
`iscrowd` (int)
```

--------------------------------

### __post_init__()

Source: https://obss.github.io/sahi/annotation

Validates bounding box coordinates and shift amount upon initialization.

```APIDOC
## __post_init__()

### Description
Validates bounding box coordinates and shift amount.

### Method
`__post_init__(self) -> None`

### Parameters
None

### Returns
None

### Raises
ValueError: If box does not contain 4 non-negative floats or if shift_amount does not contain 2 integers.
```

--------------------------------

### Postprocessor Initialization

Source: https://obss.github.io/sahi/postprocess/backends

Initializes the postprocessor with configuration parameters for matching predictions.

```APIDOC
## __init__(match_threshold=0.5, match_metric='IOU', class_agnostic=True)

### Description
Initialize the postprocessor with configuration parameters.

### Parameters
#### Path Parameters
- **match_threshold** (float) - Optional - Minimum overlap value (IoU or IoS) to consider two predictions as matching. Default: 0.5
- **match_metric** (str) - Optional - Overlap metric, "IOU" or "IOS". Default: 'IOU'
- **class_agnostic** (bool) - Optional - If True, apply postprocessing across all categories. If False, apply per category independently. Default: True
```

--------------------------------

### CocoAnnotation.category_name

Source: https://obss.github.io/sahi/api

Gets or sets the category name of the annotation.

```APIDOC
## CocoAnnotation.category_name

### Description
Gets or sets the category name of the annotation.

### Property
`category_name` (str | None)

### Setter
`category_name(self, n: str) -> None`

### Setter Parameters
#### Arguments
- **n** (str) - The new category name. Must be a string.
```

--------------------------------

### CocoAnnotation.category_id

Source: https://obss.github.io/sahi/api

Gets or sets the category ID of the annotation.

```APIDOC
## CocoAnnotation.category_id

### Description
Gets or sets the category id of the annotation.

### Property
`category_id` (int)

### Setter
`category_id(self, i: int) -> None`

### Setter Parameters
#### Arguments
- **i** (int) - The new category ID. Must be an integer.
```

--------------------------------

### ObjectPredictionList To Tensor

Source: https://obss.github.io/sahi/api

Convert to torch.Tensor. Requires torch to be installed.

```APIDOC
## totensor 

### Description
Convert to torch.Tensor. Requires torch to be installed.

### Method
totensor

### Endpoint
None

### Parameters
#### Path Parameters
None

#### Query Parameters
None

#### Request Body
None

### Returns
object: A torch.Tensor representation of the predictions.
```

--------------------------------

### Yolov5DetectionModel Initialization

Source: https://obss.github.io/sahi/models/yolov5

Initializes the YOLOv5 detection model, ensuring necessary packages like 'yolov5' and 'torch' are available.

```APIDOC
## Yolov5DetectionModel.__init__

### Description
Initializes the YOLOv5 detection model.

### Parameters
* `*args`: Variable length argument list.
* `**kwargs`: Arbitrary keyword arguments.

### Returns
None
```

--------------------------------

### DetectionModel.check_dependencies

Source: https://obss.github.io/sahi/api

Checks if all necessary dependencies for the current model are installed.

```APIDOC
## DetectionModel.check_dependencies

### Description
Verifies that all required libraries and packages for the loaded detection model are installed and accessible.

### Method
`check_dependencies`

### Parameters
#### Path Parameters
None

#### Query Parameters
None

#### Request Body
None

### Request Example
```python
from sahi.auto_model import AutoDetectionModel

# Load your detection model
detection_model = AutoDetectionModel.from_pretrained(
    model_type='detectron2',
    model_path='path/to/model.pkl',
    image_size=640
)

# Check dependencies
detection_model.check_dependencies()
```

### Response
#### Success Response (200)
None. If all dependencies are met, the function completes without error.

#### Response Example
None

#### Error Response
Raises an `ImportError` or similar exception if a required dependency is missing.
```

--------------------------------

### totensor

Source: https://obss.github.io/sahi/postprocess/utils

Converts the ObjectPredictionList to a torch.Tensor. Requires torch to be installed.

```APIDOC
## totensor()

### Description
Convert to torch.Tensor. Requires torch to be installed.

### Returns
- **object** - A torch.Tensor representation of the predictions.

### Source Code
```python
def totensor(self) -> object:
    """Convert to torch.Tensor. Requires torch to be installed."""
    return object_prediction_list_to_torch(self)
```
```

--------------------------------

### ObjectPredictionList.__init__

Source: https://obss.github.io/sahi/api

Initialize the ObjectPredictionList with a list of object predictions.

```APIDOC
## __init__(prediction_list)

### Description
Initialize with a list of object predictions.

### Parameters
#### Path Parameters
- **prediction_list** (list) - Required - List of ObjectPrediction instances.

### Returns
None
```

--------------------------------

### Initialize TorchVisionDetectionModel

Source: https://obss.github.io/sahi/models/torchvision

Initializes the TorchVision detection model, ensuring 'torch' and 'torchvision' are listed as required packages.

```python
class TorchVisionDetectionModel(DetectionModel):
    """TorchVision object detection model.

    Supports various TorchVision detection models like Faster R-CNN, Mask R-CNN, etc.
    """

    def __init__(self, *args: object, **kwargs: object) -> None:
        """Initialize TorchVision detection model."""
        existing_packages = getattr(self, "required_packages", None) or []
        self.required_packages = [*list(existing_packages), "torch", "torchvision"]
        super().__init__(*args, **kwargs)  # type: ignore[misc, arg-type]
```

--------------------------------

### ShapelyAnnotation.area

Source: https://obss.github.io/sahi/api

Gets the total area of all polygons within the annotation.

```APIDOC
## ShapelyAnnotation.area (property)

### Description
Get the total area of all polygons.

### Returns
- **int** - The total area of all polygons.

```

--------------------------------

### CocoAnnotation.image_id

Source: https://obss.github.io/sahi/api

Gets or sets the image ID associated with the annotation.

```APIDOC
## CocoAnnotation.image_id

### Description
Gets or sets the image id of the annotation.

### Property
`image_id` (int | None)

### Setter
`image_id(self, i: int) -> None`

### Setter Parameters
#### Arguments
- **i** (int) - The new image ID. Must be an integer.
```

--------------------------------

### print_environment_info

Source: https://obss.github.io/sahi/api

Logs the version information for all commonly used SAHI dependency packages.

```APIDOC
## print_environment_info()

### Description
Log version info for all commonly used SAHI dependency packages.
```

--------------------------------

### __init__

Source: https://obss.github.io/sahi/prediction

Initializes a PredictionResult object with detected objects, the source image, and optional duration information.

```APIDOC
## __init__

### Description
Initialize a PredictionResult.

### Parameters
#### Path Parameters
- **object_prediction_list** (list[ObjectPrediction]) - Required - list[ObjectPrediction] Detected objects for this image.
- **image** (Image.Image | str | ndarray) - Required - Image.Image or str or np.ndarray The source image as a PIL Image, file path, or numpy array.
#### Query Parameters
- **durations_in_seconds** (dict[str, Any]) - Optional - dict[str, Any] Elapsed times for profiling (e.g. inference, postprocess). Default: dict()
```

--------------------------------

### HuggingfaceDetectionModel.num_categories

Source: https://obss.github.io/sahi/api

Property to get the number of categories the model can detect.

```APIDOC
## num_categories

### Description
Returns the number of categories the model is trained to detect.

### Returns
- **num_categories** (int) - The total number of detectable categories.
```