MeanAbsoluteError¶

class mmeval.metrics.MeanAbsoluteError(**kwargs)[source]¶

Mean Absolute Error metric for image.

Formula: mean(abs(a-b)).

Parameters: **kwargs – Keyword parameters passed to BaseMetric.

Examples

>>> from mmeval import MeanAbsoluteError as MAE
>>> import numpy as np
>>>
>>> mae = MAE()
>>> gts = np.random.randint(0, 255, size=(3, 32, 32))
>>> preds = np.random.randint(0, 255, size=(3, 32, 32))
>>> mae(preds, gts)  
{'mae': ...}

Calculate MeanAbsoluteError between 2 images with mask:

>>> img1 = np.ones((32, 32, 3))
>>> img2 = np.ones((32, 32, 3)) * 2
>>> mask = np.ones((32, 32, 3)) * 2
>>> mask[:16] *= 0
>>> MAE.compute_mae(img1, img2, mask)
0.003921568627

add(predictions: Sequence[numpy.ndarray], groundtruths: Sequence[numpy.ndarray], masks: Optional[Sequence[numpy.ndarray]] = None) → None[source]¶

Add MeanAbsoluteError score of batch to self._results

Parameters

predictions (Sequence[np.ndarray]) – Predictions of the model.
groundtruths (Sequence[np.ndarray]) – The ground truth images.
masks (Sequence[np.ndarray], optional) – Mask images. Defaults to None.

static compute_mae(prediction: numpy.ndarray, groundtruth: numpy.ndarray, mask: Optional[numpy.ndarray] = None) → numpy.float32[source]¶

Calculate MeanAbsoluteError (Mean Absolute Error).

Parameters

prediction (np.ndarray) – Images with range [0, 255].
groundtruth (np.ndarray) – Images with range [0, 255].
mask (np.ndarray, optional) – Mask of evaluation.

Returns

MeanAbsoluteError result.

Return type

np.float32

compute_metric(results: List[numpy.float32]) → Dict[str, float][source]¶

Compute the MeanAbsoluteError metric.

This method would be invoked in BaseMetric.compute after distributed synchronization.

Parameters: results (List[np.float32]) – A list that consisting the MeanAbsoluteError score. This list has already been synced across all ranks.
Returns: The computed MeanAbsoluteError metric.
Return type: Dict[str, float]