def _create_octconv_residual_block(inputs, ch, N, alpha):
high, low = inputs
# OctConv with skip connections
for i in range(N):
# adjust channels
if i == 0:
skip_high = layers.Conv2D(int(ch * (1 - alpha)), 1)(high)
skip_high = layers.BatchNormalization()(skip_high)
skip_high = layers.Activation("relu")(skip_high)
skip_low = layers.Conv2D(int(ch * alpha), 1)(low)
skip_low = layers.BatchNormalization()(skip_low)
skip_low = layers.Activation("relu")(skip_low)
else:
skip_high, skip_low = high, low
high, low = oct_conv.OctConv2D(filters=ch, alpha=alpha)([high, low])
high = layers.BatchNormalization()(high)
high = layers.Activation("relu")(high)
low = layers.BatchNormalization()(low)
low = layers.Activation("relu")(low)
high, low = oct_conv.OctConv2D(filters=ch, alpha=alpha)([high, low])
high = layers.BatchNormalization()(high)
high = layers.Activation("relu")(high)
low = layers.BatchNormalization()(low)
low = layers.Activation("relu")(low)
high = layers.Add()([high, skip_high])
low = layers.Add()([low, skip_low])
return [high, low]
def create_octconvmlp_wide_resnet(alpha, N=4, k=10):
"""
Create OctConvmlp Wide ResNet(N=4, k=10)
"""
# Input
input = layers.Input((32, 32, 3))
# downsampling for lower
low = layers.AveragePooling2D(2)(input)
# 16 channels block
high, low = oct_conv.OctConv2D(filters=16, alpha=alpha)([input, low])
high = layers.Conv2D(int(16 * (1 - alpha)), 1)(high)
low = layers.Conv2D(int(16 * alpha), 1)(low)
high = layers.Conv2D(int(16 * (1 - alpha)), 1)(high)
low = layers.Conv2D(int(16 * alpha), 1)(low)
high = layers.BatchNormalization()(high)
high = layers.Activation("relu")(high)
low = layers.BatchNormalization()(low)
low = layers.Activation("relu")(low)
# 1st block
high, low = _create_octconvmlp_residual_block([high, low], 16 * k, N,
alpha)
# 2nd block
high = layers.AveragePooling2D(2)(high)
low = layers.AveragePooling2D(2)(low)
high, low = _create_octconvmlp_residual_block([high, low], 32 * k, N,
alpha)
# 3rd block
high = layers.AveragePooling2D(2)(high)
low = layers.AveragePooling2D(2)(low)
high, low = _create_octconvmlp_residual_block([high, low], 64 * k, N,
alpha)
# concat
high = layers.AveragePooling2D(2)(high)
x = layers.Concatenate()([high, low])
x = layers.Conv2D(64 * k, 1)(x)
x = layers.BatchNormalization()(x)
x = layers.Activation("relu")(x)
# FC
x = layers.GlobalAveragePooling2D()(x)
x = layers.Dense(10, activation="softmax")(x)
model = Model(input, x)
return model
def create_octcovmlp_vgg(alpha):
# input
input = layers.Input((32, 32, 3))
# downsampling for lower
low = layers.AveragePooling2D(2)(input)
# Block 1
high, low = oct_conv.OctConv2D(filters=64, alpha=alpha)([input, low])
high = layers.Conv2D(int(64 * (1 - alpha)), 1)(high)
low = layers.Conv2D(int(64 * alpha), 1)(low)
high = layers.Conv2D(int(64 * (1 - alpha)), 1)(high)
low = layers.Conv2D(int(64 * alpha), 1)(low)
high = layers.BatchNormalization()(high)
high = layers.Activation("relu")(high)
low = layers.BatchNormalization()(low)
low = layers.Activation("relu")(low)
high, low = oct_conv.OctConv2D(filters=64, alpha=alpha)([high, low])
high = layers.Conv2D(int(64 * (1 - alpha)), 1)(high)
low = layers.Conv2D(int(64 * alpha), 1)(low)
high = layers.Conv2D(int(64 * (1 - alpha)), 1)(high)
low = layers.Conv2D(int(64 * alpha), 1)(low)
high = layers.BatchNormalization()(high)
high = layers.Activation("relu")(high)
low = layers.BatchNormalization()(low)
low = layers.Activation("relu")(low)
# Block 2
high, low = oct_conv.OctConv2D(filters=128, alpha=alpha)([high, low])
high = layers.Conv2D(int(128 * (1 - alpha)), 1)(high)
low = layers.Conv2D(int(128 * alpha), 1)(low)
high = layers.Conv2D(int(128 * (1 - alpha)), 1)(high)
low = layers.Conv2D(int(128 * alpha), 1)(low)
high = layers.BatchNormalization()(high)
high = layers.Activation("relu")(high)
low = layers.BatchNormalization()(low)
low = layers.Activation("relu")(low)
high, low = oct_conv.OctConv2D(filters=128, alpha=alpha)([high, low])
high = layers.Conv2D(int(128 * (1 - alpha)), 1)(high)
low = layers.Conv2D(int(128 * alpha), 1)(low)
high = layers.Conv2D(int(128 * (1 - alpha)), 1)(high)
low = layers.Conv2D(int(128 * alpha), 1)(low)
high = layers.BatchNormalization()(high)
high = layers.Activation("relu")(high)
low = layers.BatchNormalization()(low)
low = layers.Activation("relu")(low)
high = layers.MaxPooling2D((2, 2), strides=(2, 2))(high)
low = layers.MaxPooling2D((2, 2), strides=(2, 2))(low)
# Block 3
high, low = oct_conv.OctConv2D(filters=256, alpha=alpha)([high, low])
high = layers.Conv2D(int(256 * (1 - alpha)), 1)(high)
low = layers.Conv2D(int(256 * alpha), 1)(low)
high = layers.Conv2D(int(256 * (1 - alpha)), 1)(high)
low = layers.Conv2D(int(256 * alpha), 1)(low)
high = layers.BatchNormalization()(high)
high = layers.Activation("relu")(high)
low = layers.BatchNormalization()(low)
low = layers.Activation("relu")(low)
high, low = oct_conv.OctConv2D(filters=256, alpha=alpha)([high, low])
high = layers.Conv2D(int(256 * (1 - alpha)), 1)(high)
low = layers.Conv2D(int(256 * alpha), 1)(low)
high = layers.Conv2D(int(256 * (1 - alpha)), 1)(high)
low = layers.Conv2D(int(256 * alpha), 1)(low)
high = layers.BatchNormalization()(high)
high = layers.Activation("relu")(high)
low = layers.BatchNormalization()(low)
low = layers.Activation("relu")(low)
high, low = oct_conv.OctConv2D(filters=256, alpha=alpha)([high, low])
high = layers.Conv2D(int(256 * (1 - alpha)), 1)(high)
low = layers.Conv2D(int(256 * alpha), 1)(low)
high = layers.Conv2D(int(256 * (1 - alpha)), 1)(high)
low = layers.Conv2D(int(256 * alpha), 1)(low)
high = layers.BatchNormalization()(high)
high = layers.Activation("relu")(high)
low = layers.BatchNormalization()(low)
low = layers.Activation("relu")(low)
# Block 4
high, low = oct_conv.OctConv2D(filters=512, alpha=alpha)([high, low])
high = layers.Conv2D(int(512 * (1 - alpha)), 1)(high)
low = layers.Conv2D(int(512 * alpha), 1)(low)
high = layers.Conv2D(int(512 * (1 - alpha)), 1)(high)
low = layers.Conv2D(int(512 * alpha), 1)(low)
high = layers.BatchNormalization()(high)
high = layers.Activation("relu")(high)
low = layers.BatchNormalization()(low)
low = layers.Activation("relu")(low)
high, low = oct_conv.OctConv2D(filters=512, alpha=alpha)([high, low])
high = layers.Conv2D(int(512 * (1 - alpha)), 1)(high)
low = layers.Conv2D(int(512 * alpha), 1)(low)
high = layers.Conv2D(int(512 * (1 - alpha)), 1)(high)
low = layers.Conv2D(int(512 * alpha), 1)(low)
high = layers.BatchNormalization()(high)
high = layers.Activation("relu")(high)
low = layers.BatchNormalization()(low)
low = layers.Activation("relu")(low)
high, low = oct_conv.OctConv2D(filters=512, alpha=alpha)([high, low])
high = layers.Conv2D(int(512 * (1 - alpha)), 1)(high)
low = layers.Conv2D(int(512 * alpha), 1)(low)
high = layers.Conv2D(int(512 * (1 - alpha)), 1)(high)
low = layers.Conv2D(int(512 * alpha), 1)(low)
high = layers.BatchNormalization()(high)
high = layers.Activation("relu")(high)
low = layers.BatchNormalization()(low)
low = layers.Activation("relu")(low)
high = layers.MaxPooling2D((2, 2), strides=(2, 2))(high)
low = layers.MaxPooling2D((2, 2), strides=(2, 2))(low)
# Block 5
high, low = oct_conv.OctConv2D(filters=512, alpha=alpha)([high, low])
high = layers.Conv2D(int(512 * (1 - alpha)), 1)(high)
low = layers.Conv2D(int(512 * alpha), 1)(low)
high = layers.Conv2D(int(512 * (1 - alpha)), 1)(high)
low = layers.Conv2D(int(512 * alpha), 1)(low)
high = layers.BatchNormalization()(high)
high = layers.Activation("relu")(high)
low = layers.BatchNormalization()(low)
low = layers.Activation("relu")(low)
high, low = oct_conv.OctConv2D(filters=512, alpha=alpha)([high, low])
high = layers.Conv2D(int(512 * (1 - alpha)), 1)(high)
low = layers.Conv2D(int(512 * alpha), 1)(low)
high = layers.Conv2D(int(512 * (1 - alpha)), 1)(high)
low = layers.Conv2D(int(512 * alpha), 1)(low)
high = layers.BatchNormalization()(high)
high = layers.Activation("relu")(high)
low = layers.BatchNormalization()(low)
low = layers.Activation("relu")(low)
high, low = oct_conv.OctConv2D(filters=512, alpha=alpha)([high, low])
high = layers.Conv2D(int(512 * (1 - alpha)), 1)(high)
low = layers.Conv2D(int(512 * alpha), 1)(low)
high = layers.Conv2D(int(512 * (1 - alpha)), 1)(high)
low = layers.Conv2D(int(512 * alpha), 1)(low)
high = layers.BatchNormalization()(high)
high = layers.Activation("relu")(high)
low = layers.BatchNormalization()(low)
low = layers.Activation("relu")(low)
# concat
high = layers.AveragePooling2D(2)(high)
x = layers.Concatenate()([high, low])
x = layers.Conv2D(4096, 1)(x)
x = layers.BatchNormalization()(x)
x = layers.Activation("relu")(x)
# FC
x = layers.GlobalAveragePooling2D()(x)
x = layers.Dense(10, activation="softmax")(x)
model = Model(input, x)
return model