def evaluate(self):
"""Compute evaluation result.
Returns:
A named tuple with the following fields -
average_precision: float numpy array of average precision for
each class.
mean_ap: mean average precision of all classes, float scalar
precisions: List of precisions, each precision is a float numpy
array
recalls: List of recalls, each recall is a float numpy array
corloc: numpy float array
mean_corloc: Mean CorLoc score for each class, float scalar
"""
if (self.num_gt_instances_per_class == 0).any():
logging.warn(
'The following classes have no ground truth examples: %s',
np.squeeze(np.argwhere(self.num_gt_instances_per_class == 0)) +
self.label_id_offset)
if self.use_weighted_mean_ap:
all_scores = np.array([], dtype=float)
all_tp_fp_labels = np.array([], dtype=bool)
for class_index in range(self.num_class):
if self.num_gt_instances_per_class[class_index] == 0:
continue
if not self.scores_per_class[class_index]:
scores = np.array([], dtype=float)
tp_fp_labels = np.array([], dtype=float)
else:
scores = np.concatenate(self.scores_per_class[class_index])
tp_fp_labels = np.concatenate(self.tp_fp_labels_per_class[class_index])
if self.use_weighted_mean_ap:
all_scores = np.append(all_scores, scores)
all_tp_fp_labels = np.append(all_tp_fp_labels, tp_fp_labels)
logging.info('Scores and tpfp per class label: %d', class_index)
logging.info(tp_fp_labels)
logging.info(scores)
precision, recall = metrics.compute_precision_recall(
scores, tp_fp_labels, self.num_gt_instances_per_class[class_index])
self.precisions_per_class.append(precision)
self.recalls_per_class.append(recall)
average_precision = metrics.compute_average_precision(precision, recall)
self.average_precision_per_class[class_index] = average_precision
self.corloc_per_class = metrics.compute_cor_loc(
self.num_gt_imgs_per_class,
self.num_images_correctly_detected_per_class)
if self.use_weighted_mean_ap:
num_gt_instances = np.sum(self.num_gt_instances_per_class)
precision, recall = metrics.compute_precision_recall(
all_scores, all_tp_fp_labels, num_gt_instances)
mean_ap = metrics.compute_average_precision(precision, recall)
else:
mean_ap = np.nanmean(self.average_precision_per_class)
mean_corloc = np.nanmean(self.corloc_per_class)
return ObjectDetectionEvalMetrics(
self.average_precision_per_class, mean_ap, self.precisions_per_class,
self.recalls_per_class, self.corloc_per_class, mean_corloc)
def evaluate(self):
"""Computes evaluation result.
Returns:
A named tuple with the following fields -
average_precision: a float number corresponding to average precision.
precisions: an array of precisions.
recalls: an array of recalls.
recall@50: recall computed on 50 top-scoring samples.
recall@100: recall computed on 100 top-scoring samples.
median_rank@50: median rank computed on 50 top-scoring samples.
median_rank@100: median rank computed on 100 top-scoring samples.
"""
if self._num_gt_instances == 0:
logging.warn('No ground truth instances')
if not self._scores:
scores = np.array([], dtype=float)
tp_fp_labels = np.array([], dtype=bool)
else:
scores = np.concatenate(self._scores)
tp_fp_labels = np.concatenate(self._tp_fp_labels)
relation_field_values = np.concatenate(self._relation_field_values)
for relation_field_value, _ in (
self._num_gt_instances_per_relationship.iteritems()):
precisions, recalls = metrics.compute_precision_recall(
scores[relation_field_values == relation_field_value],
tp_fp_labels[relation_field_values == relation_field_value],
self._num_gt_instances_per_relationship[relation_field_value])
self._average_precisions[
relation_field_value] = metrics.compute_average_precision(
precisions, recalls)
self._mean_average_precision = np.mean(
self._average_precisions.values())
self._precisions, self._recalls = metrics.compute_precision_recall(
scores, tp_fp_labels, self._num_gt_instances)
self._weighted_average_precision = metrics.compute_average_precision(
self._precisions, self._recalls)
self._recall_50 = (metrics.compute_recall_at_k(self._tp_fp_labels,
self._num_gt_instances,
50))
self._median_rank_50 = (metrics.compute_median_rank_at_k(
self._tp_fp_labels, 50))
self._recall_100 = (metrics.compute_recall_at_k(
self._tp_fp_labels, self._num_gt_instances, 100))
self._median_rank_100 = (metrics.compute_median_rank_at_k(
self._tp_fp_labels, 100))
return VRDDetectionEvalMetrics(
self._weighted_average_precision, self._mean_average_precision,
self._average_precisions, self._precisions, self._recalls,
self._recall_50, self._recall_100, self._median_rank_50,
self._median_rank_100)
def test_compute_precision_recall(self):
num_gt = 10
scores = np.array([0.4, 0.3, 0.6, 0.2, 0.7, 0.1], dtype=float)
labels = np.array([0, 1, 1, 0, 0, 1], dtype=bool)
labels_float_type = np.array([0, 1, 1, 0, 0, 1], dtype=float)
accumulated_tp_count = np.array([0, 1, 1, 2, 2, 3], dtype=float)
expected_precision = accumulated_tp_count / np.array(
[1, 2, 3, 4, 5, 6])
expected_recall = accumulated_tp_count / num_gt
precision, recall = metrics.compute_precision_recall(
scores, labels, num_gt)
precision_float_type, recall_float_type = metrics.compute_precision_recall(
scores, labels_float_type, num_gt)
self.assertAllClose(precision, expected_precision)
self.assertAllClose(recall, expected_recall)
self.assertAllClose(precision_float_type, expected_precision)
self.assertAllClose(recall_float_type, expected_recall)
def test_compute_precision_recall_and_ap_no_groundtruth(self):
num_gt = 0
scores = np.array([0.4, 0.3, 0.6, 0.2, 0.7, 0.1], dtype=float)
labels = np.array([0, 0, 0, 0, 0, 0], dtype=bool)
expected_precision = None
expected_recall = None
precision, recall = metrics.compute_precision_recall(
scores, labels, num_gt)
self.assertEquals(precision, expected_precision)
self.assertEquals(recall, expected_recall)
ap = metrics.compute_average_precision(precision, recall)
self.assertTrue(np.isnan(ap))
def test_compute_precision_recall_float(self):
num_gt = 10
scores = np.array([0.4, 0.3, 0.6, 0.2, 0.7, 0.1], dtype=float)
labels_float = np.array([0, 1, 1, 0.5, 0, 1], dtype=float)
expected_precision = np.array(
[0., 0.5, 0.33333333, 0.5, 0.55555556, 0.63636364], dtype=float)
expected_recall = np.array([0., 0.1, 0.1, 0.2, 0.25, 0.35],
dtype=float)
precision, recall = metrics.compute_precision_recall(
scores, labels_float, num_gt)
self.assertAllClose(precision, expected_precision)
self.assertAllClose(recall, expected_recall)