import math
import tensorflow as tf
from bob.learn.tensorflow.metrics.embedding_accuracy import accuracy_from_embeddings
from .embedding_validation import EmbeddingValidation
class ArcFaceModel(EmbeddingValidation):
[docs] def train_step(self, data):
X, y = data
with tf.GradientTape() as tape:
logits, _ = self((X, y), training=True)
loss = self.compiled_loss(
y, logits, sample_weight=None, regularization_losses=self.losses
)
reg_loss = tf.reduce_sum(self.losses)
total_loss = loss + reg_loss
trainable_vars = self.trainable_variables
self.optimizer.minimize(total_loss, trainable_vars, tape=tape)
self.compiled_metrics.update_state(y, logits, sample_weight=None)
tf.summary.scalar("arc_face_loss", data=loss, step=self._train_counter)
tf.summary.scalar("total_loss", data=total_loss, step=self._train_counter)
self.train_loss(loss)
return {m.name: m.result() for m in self.metrics + [self.train_loss]}
[docs] def test_step(self, data):
"""
Test Step
"""
images, labels = data
# No worries, labels not used in validation
_, embeddings = self((images, labels), training=False)
self.validation_acc(accuracy_from_embeddings(labels, embeddings))
return {m.name: m.result() for m in [self.validation_acc]}
class ArcFaceLayer(tf.keras.layers.Layer):
"""
Implements the ArcFace from equation (3) of `ArcFace: Additive Angular Margin Loss for Deep Face Recognition <https://arxiv.org/abs/1801.07698>`_
Defined as:
:math:`s(cos(\\theta_i) + m`
Parameters
----------
n_classes: int
Number of classes
m: float
Margin
s: int
Scale
arc: bool
If `True`, uses arcface loss. If `False`, it's a regular dense layer
"""
def __init__(self, n_classes=10, s=30, m=0.5, arc=True):
super(ArcFaceLayer, self).__init__(name="arc_face_logits")
self.n_classes = n_classes
self.s = s
self.arc = arc
self.m = m
[docs] def build(self, input_shape):
super(ArcFaceLayer, self).build(input_shape[0])
shape = [input_shape[-1], self.n_classes]
self.W = self.add_variable("W", shape=shape)
self.cos_m = tf.identity(math.cos(self.m), name="cos_m")
self.sin_m = tf.identity(math.sin(self.m), name="sin_m")
self.th = tf.identity(math.cos(math.pi - self.m), name="th")
self.mm = tf.identity(math.sin(math.pi - self.m) * self.m)
[docs] def call(self, X, y, training=None):
if self.arc:
# normalize feature
X = tf.nn.l2_normalize(X, axis=1)
W = tf.nn.l2_normalize(self.W, axis=0)
# cos between X and W
cos_yi = tf.matmul(X, W)
# sin_yi = tf.math.sqrt(1-cos_yi**2)
sin_yi = tf.clip_by_value(tf.math.sqrt(1 - cos_yi ** 2), 0, 1)
# cos(x+m) = cos(x)*cos(m) - sin(x)*sin(m)
cos_yi_m = cos_yi * self.cos_m - sin_yi * self.sin_m
cos_yi_m = tf.where(cos_yi > self.th, cos_yi_m, cos_yi - self.mm)
# Preparing the hot-output
one_hot = tf.one_hot(
tf.cast(y, tf.int32), depth=self.n_classes, name="one_hot_mask"
)
logits = (one_hot * cos_yi_m) + ((1.0 - one_hot) * cos_yi)
logits = self.s * logits
else:
logits = tf.matmul(X, self.W)
return logits
class ArcFaceLayer3Penalties(tf.keras.layers.Layer):
"""
Implements the ArcFace loss from equation (4) of `ArcFace: Additive Angular Margin Loss for Deep Face Recognition <https://arxiv.org/abs/1801.07698>`_
Defined as:
:math:`s(cos(m_1\\theta_i + m_2) -m_3`
"""
def __init__(self, n_classes=10, s=30, m1=0.5, m2=0.5, m3=0.5):
super(ArcFaceLayer3Penalties, self).__init__(name="arc_face_logits")
self.n_classes = n_classes
self.s = s
self.m1 = m1
self.m2 = m2
self.m3 = m3
[docs] def build(self, input_shape):
super(ArcFaceLayer3Penalties, self).build(input_shape[0])
shape = [input_shape[-1], self.n_classes]
self.W = self.add_variable("W", shape=shape)
[docs] def call(self, X, y, training=None):
# normalize feature
X = tf.nn.l2_normalize(X, axis=1)
W = tf.nn.l2_normalize(self.W, axis=0)
# cos between X and W
cos_yi = tf.matmul(X, W)
# Getting the angle
theta = tf.math.acos(cos_yi)
theta = tf.clip_by_value(
theta, -1.0 + tf.keras.backend.epsilon(), 1 - tf.keras.backend.epsilon()
)
cos_yi_m = tf.math.cos(self.m1 * theta + self.m2) - self.m3
# logits = self.s*cos_theta_m
# Preparing the hot-output
one_hot = tf.one_hot(
tf.cast(y, tf.int32), depth=self.n_classes, name="one_hot_mask"
)
one_hot = tf.cast(one_hot, cos_yi_m.dtype)
logits = (one_hot * cos_yi_m) + ((1.0 - one_hot) * cos_yi)
logits = self.s * logits
return logits