def _build_loss(self):
logits = super().get_logits()
labels = super().get_labels()
num_classes = super().get_num_classes()
num_positives = None
if self._normalize_by_positives:
positives = tf.cast(tf.not_equal(labels, 0),
tf.float32) # [BATCH_SIZE, ...]
positives_dim_n = len(positives.get_shape())
axis = list(range(1, positives_dim_n))
num_positives = tf.reduce_sum(
positives, axis=axis) # [BATCH_SIZE, N_POSITIVES]
focal_loss = Loss.focal_loss(logits=logits,
labels=labels,
num_classes=num_classes,
num_positives=num_positives,
focal_gamma=self._focal_gamma)
if not self._normalize_by_positives:
focal_loss = focal_loss / float(super().get_batch_size())
super().track_loss(focal_loss, FocalTrainer.FOCAL_LOSS)
return focal_loss
def _build_loss(self):
logits = super().get_logits()
labels = super().get_labels()
num_classes = super().get_num_classes()
positives = tf.not_equal(labels, 0) # [BATCH_SIZE, ...]
positives_dim_n = len(positives.get_shape())
axis = list(range(1, positives_dim_n))
num_positives = tf.reduce_sum(positives,
axis=axis) # [BATCH_SIZE, N_POSITIVES]
focal_loss = Loss.focal_loss(logits=logits,
labels=labels,
num_classes=num_classes,
num_positives=num_positives,
focal_gamma=self._focal_gamma,
raw_tensor=True)
weights = super().get_weight_map()
focal_loss = tf.reduce_sum(focal_loss * weights)
super().track_loss(focal_loss, WeightedFocalTrainer.FOCAL_LOSS)
return focal_loss