get_kind_of_label_match(
pred_label_list,
gt_label_list,
hide_gt_over_thresh=True,
iou_threshold=0.5,
)
Creates a list of InstanceLabels for prediction labels and ground truth labels.
Based on the iou of the prediction label (mask or bbox) and
the ground truth label (=gt; mask or bbox), the labels are given a color and
are active or not. Red = gt label with no matching prediction label;
Blue = prediction label with no matching gt label; Green = prediction label which
matched at least one gt label in position and class; Yellow = prediction label
which matched at least one gt label in position but not in class.
Parameters:
-
pred_label_list
(List[InstanceLabel])
–
label list of the prediction labels
-
gt_label_list
(List[InstanceLabel])
–
label list of the ground truth labels
-
hide_gt_over_thresh
(bool, default:
True
)
–
flag to hide the ground truth labels if the ground truth label
have an iou >= iou_threshold with a prediction label
-
iou_threshold
(float, default:
0.5
)
–
threshold to determine kind of match by iou
Returns:
Source code in niceml/utilities/instancelabelmatching.py
| def get_kind_of_label_match(
pred_label_list: List[InstanceLabel],
gt_label_list: List[InstanceLabel],
hide_gt_over_thresh: bool = True,
iou_threshold: float = 0.5,
) -> Tuple[List[InstanceLabel], List[InstanceLabel]]:
"""
Creates a list of InstanceLabels for prediction labels and ground truth labels.
Based on the iou of the prediction label (mask or bbox) and
the ground truth label (=gt; mask or bbox), the labels are given a color and
are active or not. Red = gt label with no matching prediction label;
Blue = prediction label with no matching gt label; Green = prediction label which
matched at least one gt label in position and class; Yellow = prediction label
which matched at least one gt label in position but not in class.
Args:
pred_label_list: label list of the prediction labels
gt_label_list: label list of the ground truth labels
hide_gt_over_thresh: flag to hide the ground truth labels if the ground truth label
have an iou >= `iou_threshold` with a prediction label
iou_threshold: threshold to determine kind of match by iou
Returns:
List of InstanceLabel for prediction labels and ground truth labels
"""
if len(pred_label_list) == 0:
for gt_label in gt_label_list:
gt_label.active = True
gt_label.color = Color.RED
else:
for pred_label in pred_label_list:
if len(gt_label_list) == 0:
pred_label.color = Color.BLUE
pred_label.active = True
for gt_label in gt_label_list:
iou = pred_label.calc_iou(other=gt_label)
pred_label, gt_label = get_kind_of_instance_label_match(
pred_label=pred_label,
gt_label=gt_label,
iou=iou,
iou_threshold=iou_threshold,
hide_gt_over_thresh=hide_gt_over_thresh,
)
return pred_label_list, gt_label_list
|