def metriccalculation(predictions, Y_validation, numpos=None):
"""
Calculates the global average precision between the predictions and Y_validation arrays.
:param predictions:
:param Y_validation:
:param positive_labels_count: In case the Y_validation array isn't the complete, this parameter gives the
true number of positive labels.
:return: The score given by GAP.
"""
if predictions.shape != Y_validation.shape:
raise ValueError(
"Different shapes between 'predictions' and 'Y_validation'")
valcases = len(Y_validation)
gap = GAP.AveragePrecisionCalculator(20 * valcases)
predictions = np.array(predictions)
Y_validation = np.array(Y_validation)
for i in range(valcases):
p = predictions[i].argsort()[::-1]
predictions[i] = predictions[i][p]
Y_validation[i] = Y_validation[i][p]
if numpos is None:
gap.accumulate(predictions[i][:20],
Y_validation[i][:20],
num_positives=np.sum(Y_validation[i]))
else:
gap.accumulate(predictions[i][:20],
Y_validation[i][:20],
num_positives=numpos[i])
return gap.peek_ap_at_n()
def __init__(self, num_class, top_k): self.sum_hit_at_one = 0.0 self.sum_perr = 0.0 self.sum_loss = 0.0 self.map_calculator = map_calculator.MeanAveragePrecisionCalculator(num_class) self.global_ap_calculator = ap_calculator.AveragePrecisionCalculator() self.top_k = top_k self.num_examples = 0
def __init__(self, num_class):
if not isinstance(num_class, int) or num_class <= 1:
raise ValueError("num_class must be a positive integer.")
self._ap_calculators = [] # member of AveragePrecisionCalculator
self._num_class = num_class # total number of classes
for i in range(num_class):
self._ap_calculators.append(
average_precision_calculator.AveragePrecisionCalculator())
def calculate_gap(predictions, actuals, top_k=20):
"""Performs a local (numpy) calculation of the global average precision.
Only the top_k predictions are taken for each of the videos.
Args:
predictions: Matrix containing the outputs of the model.
Dimensions are 'batch' x 'num_classes'.
actuals: Matrix containing the ground truth labels.
Dimensions are 'batch' x 'num_classes'.
top_k: How many predictions to use per video.
Returns:
float: The global average precision.
"""
gap_calculator = ap_calculator.AveragePrecisionCalculator()
sparse_predictions, sparse_labels, num_positives = top_k_by_class(predictions, actuals, top_k)
gap_calculator.accumulate(flatten(sparse_predictions), flatten(sparse_labels), sum(num_positives))
return gap_calculator.peek_ap_at_n()
def __init__(self, num_class, top_k):
"""Construct an EvaluationMetrics object to store the evaluation metrics.
Args:
num_class: A positive integer specifying the number of classes.
top_k: A positive integer specifying how many predictions are considered per video.
Raises:
ValueError: An error occurred when MeanAveragePrecisionCalculator cannot
not be constructed.
"""
self.sum_hit_at_one = 0.0
self.sum_perr = 0.0
self.sum_loss = 0.0
self.map_calculator = map_calculator.MeanAveragePrecisionCalculator(num_class)
self.global_ap_calculator = ap_calculator.AveragePrecisionCalculator()
self.top_k = top_k
self.num_examples = 0
def __init__(self, num_class, filter_empty_classes=True):
"""Construct a calculator to calculate the (macro) average precision.
Args:
num_class: A positive Integer specifying the number of classes.
filter_empty_classes: whether to filter classes without any positives.
Raises:
ValueError: An error occurred when num_class is not a positive integer;
or the top_n_array is not a list of positive integers.
"""
if not isinstance(num_class, int) or num_class <= 1:
raise ValueError("num_class must be a positive integer.")
self._ap_calculators = [] # member of AveragePrecisionCalculator
self._num_class = num_class # total number of classes
self._filter_empty_classes = filter_empty_classes
for i in range(num_class):
self._ap_calculators.append(
average_precision_calculator.AveragePrecisionCalculator())
def __init__(self, num_class):
"""Construct a calculator to calculate the (macro) average precision.
Args:
num_class: A positive Integer specifying the number of classes.
top_n_array: A list of positive integers specifying the top n for each
class. The top n in each class will be used to calculate its average
precision at n.
The size of the array must be num_class.
Raises:
ValueError: An error occurred when num_class is not a positive integer;
or the top_n_array is not a list of positive integers.
"""
if not isinstance(num_class, int) or num_class <= 1:
raise ValueError("num_class must be a positive integer.")
self._ap_calculators = [] # member of AveragePrecisionCalculator
self._num_class = num_class # total number of classes
for i in range(num_class):
self._ap_calculators.append(
average_precision_calculator.AveragePrecisionCalculator())
def calculate_gap(predictions, actuals, top_k=20): gap_calculator = ap_calculator.AveragePrecisionCalculator() sparse_predictions, sparse_labels, num_positives = top_k_by_class(predictions, actuals, top_k) gap_calculator.accumulate(flatten(sparse_predictions), flatten(sparse_labels), sum(num_positives)) return gap_calculator.peek_ap_at_n()
def __init__(self, num_class, top_k):
self.top_k = top_k
self.global_ap_calculator = ap_calculator.AveragePrecisionCalculator()
