Base Metric

Base metric class for image generation evaluation.

BaseMetric

Bases: ABC

Abstract base class for all evaluation metrics.

All metrics must inherit from this class and implement the required methods.

Attributes:

Name	Type	Description
model		The generative model being evaluated.

Source code in image_gen\metrics\base.py

class BaseMetric(ABC):
    """Abstract base class for all evaluation metrics.

    All metrics must inherit from this class and implement the required methods.

    Attributes:
        model: The generative model being evaluated.
    """

    def __init__(self, model: GenerativeModel):
        """Initialize the metric with a generative model.

        Args:
            model: The generative model to be evaluated with this metric.
        """
        self.model = model

    @abstractmethod
    def __call__(self, real: Tensor, generated: Tensor, *args: Any, **kwargs: Any) -> float:
        """Compute the metric value between real and generated samples.

        Args:
            real: Tensor containing real samples.
            generated: Tensor containing generated samples.
            *args: Additional positional arguments.
            **kwargs: Additional keyword arguments.

        Returns:
            float: The computed metric value.
        """
        pass

    @property
    @abstractmethod
    def name(self) -> str:
        """Get the name of the metric.

        Returns:
            str: The name of the metric.
        """
        pass

    @property
    @abstractmethod
    def is_lower_better(self) -> bool:
        """Indicates whether a lower metric value is better.

        Returns:
            bool: True if lower values indicate better performance, False otherwise.
        """
        pass

    def config(self) -> dict:
        """Get the configuration parameters for this metric.

        Returns:
            dict: Dictionary containing configuration parameters.
        """
        return {}

    def __str__(self) -> str:
        """Get a string representation of the metric.

        Returns:
            str: String representation including class name and parameters.
        """
        config = self.config()
        params = ", ".join(f"{k}: {v}" for k, v in config.items())
        return f"{self._class_name}({params})"

    @property
    def _class_name(self) -> str:
        """Get the class name.

        This property will be automatically overridden in custom classes made by users.

        Returns:
            str: The class name.
        """
        return self.__class__.__name__

is_lower_better abstractmethod property

Indicates whether a lower metric value is better.

Returns:

Name	Type	Description
bool	bool	True if lower values indicate better performance, False otherwise.

name abstractmethod property

Get the name of the metric.

Returns:

Name	Type	Description
str	str	The name of the metric.

__call__(real, generated, *args, **kwargs) abstractmethod

Compute the metric value between real and generated samples.

Parameters:

Name	Type	Description	Default
real	Tensor	Tensor containing real samples.	required
generated	Tensor	Tensor containing generated samples.	required
*args	Any	Additional positional arguments.	()
**kwargs	Any	Additional keyword arguments.	{}

Returns:

Name	Type	Description
float	float	The computed metric value.

Source code in image_gen\metrics\base.py

@abstractmethod
def __call__(self, real: Tensor, generated: Tensor, *args: Any, **kwargs: Any) -> float:
    """Compute the metric value between real and generated samples.

    Args:
        real: Tensor containing real samples.
        generated: Tensor containing generated samples.
        *args: Additional positional arguments.
        **kwargs: Additional keyword arguments.

    Returns:
        float: The computed metric value.
    """
    pass

__init__(model)

Initialize the metric with a generative model.

Parameters:

Name	Type	Description	Default
model	GenerativeModel	The generative model to be evaluated with this metric.	required

Source code in image_gen\metrics\base.py

def __init__(self, model: GenerativeModel):
    """Initialize the metric with a generative model.

    Args:
        model: The generative model to be evaluated with this metric.
    """
    self.model = model

__str__()

Get a string representation of the metric.

Returns:

Name	Type	Description
str	str	String representation including class name and parameters.

Source code in image_gen\metrics\base.py

def __str__(self) -> str:
    """Get a string representation of the metric.

    Returns:
        str: String representation including class name and parameters.
    """
    config = self.config()
    params = ", ".join(f"{k}: {v}" for k, v in config.items())
    return f"{self._class_name}({params})"

config()

Get the configuration parameters for this metric.

Returns:

Name	Type	Description
dict	dict	Dictionary containing configuration parameters.

Source code in image_gen\metrics\base.py

def config(self) -> dict:
    """Get the configuration parameters for this metric.

    Returns:
        dict: Dictionary containing configuration parameters.
    """
    return {}