View source on GitHub

Implements the focal loss function.

Focal loss was first introduced in the RetinaNet paper ( Focal loss is extremely useful for classification when you have highly imbalanced classes. It down-weights well-classified examples and focuses on hard examples. The loss value is much high for a sample which is misclassified by the classifier as compared to the loss value corresponding to a well-classified example. One of the best use-cases of focal loss is its usage in object detection where the imbalance between the background class and other classes is extremely high.


fl = tfa.losses.SigmoidFocalCrossEntropy()
loss = fl(
  y_true = [[1.0], [1.0], [0.0]],
  y_pred = [[0.97], [0.91], [0.03]])
print('Loss: ', loss.numpy())  # Loss: [6.8532745e-06,

Usage with tf.keras API:

model = tf.keras.Model(inputs, outputs)
model.compile('sgd', loss=tfa.losses.SigmoidFocalCrossEntropy())

Args alpha: balancing factor, default value is 0.25 gamma: modulating factor, default value is 2.0

Weighted loss float Tensor. If reduction is NONE, this has the same shape as y_true; otherwise, it is scalar.

ValueError If the shape of sample_weight is invalid or value of gamma is less than zero

fn The loss function to wrap, with signature fn(y_true, y_pred, **kwargs).
reduction (Optional) Type of tf.keras.losses.Reduction to apply to loss. Default value is AUTO. AUTO indicates that the reduction option will be determined by the usage context. For almost all cases this defaults to SUM_OVER_BATCH_SIZE. When used with tf.distribute.Strategy, outside of built-in training loops such as tf.keras compile and fit, using AUTO or SUM_OVER_BATCH_SIZE will raise an error. Please see this custom training tutorial for more details.
name (Optional) name for the loss.
**kwargs The keyword arguments that are passed on to fn.



Instantiates a Loss from its config (output of get_config()).

config Output of get_config().

A Loss instance.


View source

Returns the config dictionary for a Loss instance.


Invokes the Loss instance.

y_true Ground truth values. shape = [batch_size, d0, .. dN], except sparse loss functions such as sparse categorical crossentropy where shape = [batch_size, d0, .. dN-1]
y_pred The predicted values. shape = [batch_size, d0, .. dN]
sample_weight Optional sample_weight acts as a coefficient for the loss. If a scalar is provided, then the loss is simply scaled by the given value. If sample_weight is a tensor of size [batch_size], then the total loss for each sample of the batch is rescaled by the corresponding element in the sample_weight vector. If the shape of sample_weight is [batch_size, d0, .. dN-1] (or can be broadcasted to this shape), then each loss element of y_pred is scaled by the corresponding value of sample_weight. (Note ondN-1: all loss functions reduce by 1 dimension, usually axis=-1.)

Weighted loss float Tensor. If reduction is NONE, this has shape [batch_size, d0, .. dN-1]; otherwise, it is scalar. (Note dN-1 because all loss functions reduce by 1 dimension, usually axis=-1.)

ValueError If the shape of sample_weight is invalid.