def create_pyramid_features(C3, C4, C5, feature_size=256):
''' create pyramid feature '''
# upsample C5 to get P5 from the FPN paper
P5 = keras.layers.Conv2D(feature_size,
kernel_size=1,
strides=1,
padding='same',
name='C5_reduced')(C5)
P5_upsampled = layers.UpsampleLike(name='P5_upsampled')([P5, C4])
P5 = keras.layers.Conv2D(feature_size,
kernel_size=3,
strides=1,
padding='same',
name='P5')(P5)
# add P5 elementwise to C4
P4 = keras.layers.Conv2D(feature_size,
kernel_size=1,
strides=1,
padding='same',
name='C4_reduced')(C4)
P4 = keras.layers.Add(name='P4_merged')([P5_upsampled, P4])
P4_upsampled = layers.UpsampleLike(name='P4_upsampled')([P4, C3])
P4 = keras.layers.Conv2D(feature_size,
kernel_size=3,
strides=1,
padding='same',
name='P4')(P4)
# add P4 elementwise to C3
P3 = keras.layers.Conv2D(feature_size,
kernel_size=1,
strides=1,
padding='same',
name='C3_reduced')(C3)
P3 = keras.layers.Add(name='P3_merged')([P4_upsampled, P3])
P3 = keras.layers.Conv2D(feature_size,
kernel_size=3,
strides=1,
padding='same',
name='P3')(P3)
# "P6 is obtained via a 3x3 stride-2 conv on C5"
P6 = keras.layers.Conv2D(feature_size,
kernel_size=3,
strides=2,
padding='same',
name='P6')(C5)
# "P7 is computed by applying ReLU followed by a 3x3 stride-2 conv on P6"
P7 = keras.layers.Activation('relu', name='C6_relu')(P6)
P7 = keras.layers.Conv2D(feature_size,
kernel_size=3,
strides=2,
padding='same',
name='P7')(P7)
return P3, P4, P5, P6, P7
def __create_pyramid_features(C2, C3, C4, C5, feature_size=256):
"""!@brief
Creates the FPN layers on top of the backbone features.
@param C2 : Feature stage C2 from the backbone.
@param C3 : Feature stage C3 from the backbone.
@param C4 : Feature stage C4 from the backbone.
@param C5 : Feature stage C5 from the backbone.
@param feature_size : The feature size to use for the resulting feature levels.
@return
A list of feature levels [P2, P3, P4, P5].
"""
# upsample C5 to get P5 from the FPN paper
P5 = keras.layers.Conv2D(feature_size, kernel_size=1, strides=1, padding='same', name='C5_reduced')(C5)
P5_upsampled = layers.UpsampleLike(name='P5_upsampled')([P5, C4])
P5 = keras.layers.Conv2D(feature_size, kernel_size=3, strides=1, padding='same', name='P5')(P5)
# add P5 elementwise to C4
P4 = keras.layers.Conv2D(feature_size, kernel_size=1, strides=1, padding='same', name='C4_reduced')(C4)
P4 = keras.layers.Add(name='P4_merged')([P5_upsampled, P4])
P4_upsampled = layers.UpsampleLike(name='P4_upsampled')([P4, C3])
P4 = keras.layers.Conv2D(feature_size, kernel_size=3, strides=1, padding='same', name='P4')(P4)
# add P4 elementwise to C3
P3 = keras.layers.Conv2D(feature_size, kernel_size=1, strides=1, padding='same', name='C3_reduced')(C3)
P3 = keras.layers.Add(name='P3_merged')([P4_upsampled, P3])
P3_upsampled = layers.UpsampleLike(name='P3_upsampled')([P3, C2])
P3 = keras.layers.Conv2D(feature_size, kernel_size=3, strides=1, padding='same', name='P3')(P3)
# add P3 elementwise to C2
P2 = keras.layers.Conv2D(feature_size, kernel_size=1, strides=1, padding='same', name='C2_reduced')(C2)
P2 = keras.layers.Add(name='P2_merged')([P3_upsampled, P2])
P2 = keras.layers.Conv2D(feature_size, kernel_size=3, strides=1, padding='same', name='P2')(P2)
# # "P6 is obtained via a 3x3 stride-2 conv on C5"
# P6 = keras.layers.Conv2D(feature_size, kernel_size=3, strides=2, padding='same', name='P6')(C5)
# # "P7 is computed by applying ReLU followed by a 3x3 stride-2 conv on P6"
# P7 = keras.layers.Activation('relu', name='C6_relu')(P6)
# P7 = keras.layers.Conv2D(feature_size, kernel_size=3, strides=2, padding='same', name='P7')(P7)
return [P2, P3, P4, P5]