def eval_instance_segmentation_voc(pred_masks,
pred_labels,
pred_scores,
gt_masks,
gt_labels,
iou_thresh=0.5,
use_07_metric=False):
"""Calculate average precisions based on evaluation code of PASCAL VOC.
This function evaluates predicted masks obtained from a dataset
which has :math:`N` images by using average precision for each class.
The code is based on the evaluation code used in `FCIS`_.
.. _`FCIS`: https://arxiv.org/abs/1611.07709
Args:
pred_masks (iterable of numpy.ndarray): See the table below.
pred_labels (iterable of numpy.ndarray): See the table below.
pred_scores (iterable of numpy.ndarray): See the table below.
gt_masks (iterable of numpy.ndarray): See the table below.
gt_labels (iterable of numpy.ndarray): See the table below.
iou_thresh (float): A prediction is correct if its Intersection over
Union with the ground truth is above this value.
use_07_metric (bool): Whether to use PASCAL VOC 2007 evaluation metric
for calculating average precision. The default value is
:obj:`False`.
.. csv-table::
:header: name, shape, dtype, format
:obj:`pred_masks`, ":math:`[(R, H, W)]`", :obj:`bool`, --
:obj:`pred_labels`, ":math:`[(R,)]`", :obj:`int32`, \
":math:`[0, \#fg\_class - 1]`"
:obj:`pred_scores`, ":math:`[(R,)]`", :obj:`float32`, \
--
:obj:`gt_masks`, ":math:`[(R, H, W)]`", :obj:`bool`, --
:obj:`gt_labels`, ":math:`[(R,)]`", :obj:`int32`, \
":math:`[0, \#fg\_class - 1]`"
Returns:
dict:
The keys, value-types and the description of the values are listed
below.
* **ap** (*numpy.ndarray*): An array of average precisions. \
The :math:`l`-th value corresponds to the average precision \
for class :math:`l`. If class :math:`l` does not exist in \
either :obj:`pred_labels` or :obj:`gt_labels`, the corresponding \
value is set to :obj:`numpy.nan`.
* **map** (*float*): The average of Average Precisions over classes.
"""
prec, rec = calc_instance_segmentation_voc_prec_rec(
pred_masks, pred_labels, pred_scores, gt_masks, gt_labels, iou_thresh)
ap = calc_detection_voc_ap(prec, rec, use_07_metric=use_07_metric)
return {'ap': ap, 'map': np.nanmean(ap)}
def eval_instance_segmentation_voc(pred_masks,
pred_labels,
pred_scores,
gt_masks,
gt_labels,
iou_thresh=0.5,
use_07_metric=False,
n_pos=None,
score=None,
match=None):
n_pos, score, match = cal_running_instance(pred_masks, pred_labels,
pred_scores, gt_masks,
gt_labels, iou_thresh, n_pos,
score, match)
prec, rec = calc_instance_segmentation_voc_prec_rec(n_pos, score, match)
ap = calc_detection_voc_ap(prec, rec, use_07_metric=use_07_metric)
return {
'ap': ap,
'map': np.nanmean(ap),
"n_pos": n_pos,
"score": score,
"match": match
}
def eval_instseg_voc(pred_masks,
pred_labels,
pred_scores,
gt_masks,
gt_labels,
gt_difficults=None,
iou_thresh=0.5,
use_07_metric=False):
assert not use_07_metric
prec, rec, sq, dq, pq = calc_instseg_voc_prec_rec(pred_masks,
pred_labels,
pred_scores,
gt_masks,
gt_labels,
gt_difficults,
iou_thresh=iou_thresh)
ap = calc_detection_voc_ap(prec, rec, use_07_metric=use_07_metric)
return {
'map': np.nanmean(ap),
'msq': np.nanmean(np.mean(sq, axis=1)),
'mdq': np.nanmean(dq),
'mpq': np.nanmean(pq),
'msq/vis': np.nanmean(sq[:, 0]),
'msq/occ': np.nanmean(sq[:, 1]),
}
def eval_instance_segmentation_voc(
generator, iou_threshs=(0.5, 0.7), use_07_metric=False):
prec, rec = calc_instseg_voc_prec_rec(generator, iou_threshs)
ap = {}
for iou_thresh in iou_threshs:
ap_thresh = calc_detection_voc_ap(
prec[iou_thresh], rec[iou_thresh], use_07_metric=use_07_metric)
ap['ap{}'.format(iou_thresh)] = ap_thresh
ap['map{}'.format(iou_thresh)] = np.nanmean(ap_thresh)
return ap
def eval_instseg_voc(
pred_masks, pred_labels, pred_scores, gt_masks, gt_labels,
gt_difficults=None,
iou_thresh=0.5, use_07_metric=False):
prec, rec = calc_instseg_voc_prec_rec(
pred_masks, pred_labels, pred_scores,
gt_masks, gt_labels, gt_difficults,
iou_thresh=iou_thresh)
ap = calc_detection_voc_ap(prec, rec, use_07_metric=use_07_metric)
return {'ap': ap, 'map': np.nanmean(ap)}
def test_calc_detection_voc_ap(self):
ap = calc_detection_voc_ap(
self.prec, self.rec, use_07_metric=self.use_07_metric)
np.testing.assert_almost_equal(ap, self.ap)
def test_calc_detection_voc_ap(self):
ap = calc_detection_voc_ap(self.prec,
self.rec,
use_07_metric=self.use_07_metric)
np.testing.assert_almost_equal(ap, self.ap)
def eval_multi_label_classification(pred_labels, pred_scores, gt_labels):
prec, rec = calc_multi_label_classification_prec_rec(
pred_labels, pred_scores, gt_labels)
ap = calc_detection_voc_ap(prec, rec)
return {'ap': ap, 'map': np.nanmean(ap)}
def eval_instance_segmentation_voc(pred_masks,
pred_labels,
pred_scores,
gt_masks,
gt_labels,
iou_thresh=0.5,
use_07_metric=False):
"""Calculate average precisions based on evaluation code of PASCAL VOC.
This function evaluates predicted masks obtained from a dataset
which has :math:`N` images by using average precision for each class.
The code is based on the evaluation code used in `FCIS`_.
.. _`FCIS`: https://arxiv.org/abs/1611.07709
Args:
pred_masks (iterable of numpy.ndarray): An iterable of :math:`N`
sets of masks. Its index corresponds to an index for the base
dataset. Each element of :obj:`pred_masks` is an object mask
and is an array whose shape is :math:`(R, H, W)`,
where :math:`R` corresponds
to the number of masks, which may vary among images.
pred_labels (iterable of numpy.ndarray): An iterable of labels.
Similar to :obj:`pred_masks`, its index corresponds to an
index for the base dataset. Its length is :math:`N`.
pred_scores (iterable of numpy.ndarray): An iterable of confidence
scores for predicted masks. Similar to :obj:`pred_masks`,
its index corresponds to an index for the base dataset.
Its length is :math:`N`.
gt_masks (iterable of numpy.ndarray): An iterable of ground truth
masks whose length is :math:`N`. An element of :obj:`gt_masks` is
an object mask whose shape is :math:`(R, H, W)`. Note that the
number of masks :math:`R` in each image does not need to be
same as the number of corresponding predicted masks.
gt_labels (iterable of numpy.ndarray): An iterable of ground truth
labels which are organized similarly to :obj:`gt_masks`. Its
length is :math:`N`.
iou_thresh (float): A prediction is correct if its Intersection over
Union with the ground truth is above this value.
use_07_metric (bool): Whether to use PASCAL VOC 2007 evaluation metric
for calculating average precision. The default value is
:obj:`False`.
Returns:
dict:
The keys, value-types and the description of the values are listed
below.
* **ap** (*numpy.ndarray*): An array of average precisions. \
The :math:`l`-th value corresponds to the average precision \
for class :math:`l`. If class :math:`l` does not exist in \
either :obj:`pred_labels` or :obj:`gt_labels`, the corresponding \
value is set to :obj:`numpy.nan`.
* **map** (*float*): The average of Average Precisions over classes.
"""
prec, rec = calc_instance_segmentation_voc_prec_rec(
pred_masks, pred_labels, pred_scores, gt_masks, gt_labels, iou_thresh)
ap = calc_detection_voc_ap(prec, rec, use_07_metric=use_07_metric)
return {'ap': ap, 'map': np.nanmean(ap)}
def eval_multi_label_classification(
pred_labels, pred_scores, gt_labels):
prec, rec = calc_multi_label_classification_prec_rec(
pred_labels, pred_scores, gt_labels)
ap = calc_detection_voc_ap(prec, rec)
return {'ap': ap, 'map': np.nanmean(ap)}