fastface.metric
- class fastface.metric.WiderFaceAP(iou_threshold: float = 0.5)
torchmetrics.Metric instance to calculate widerface average precision
- Parameters:
iou_threshold (float) – widerface AP score IoU threshold, default is 0.5
- class fastface.metric.AveragePrecision(iou_threshold: List | float = 0.5, area: str | None = None)
torchmetrics.Metric instance to calculate binary average precision
- Parameters:
iou_threshold (Union[List, float]) – iou threshold or list of iou thresholds
area (str) – small or medium or large or None
- Returns:
[description]
- Return type:
[type]
- class fastface.metric.AverageRecall(iou_threshold_min: float = 0.5, iou_threshold_max: float = 1.0)
torchmetrics.Metric instance to calculate average recall
\[AR = 2 \times \int_\text{iou_threshold_min}^\text{iou_threshold_max} recall(o)do\]- Parameters:
iou_threshold_min (float, optional) – minimum threshold for IoU. Defaults to 0.5.
iou_threshold_max (float, optional) – maximum threshold for IoU. Defaults to 1.0.