Styjun

How to ensure color fidelity of the same file on two computers?

How to decline a wedding invitation from a friend I haven't seen in years?

Who enforces MPAA rating adherence?

How is the excise border managed in Ireland?

Let M and N be single-digit integers. If the product 2M5 x 13N is divisible by 36, how many ordered pairs (M,N) are possible?

Has there been a multiethnic Star Trek character?

Is using 'echo' to display attacker-controlled data on the terminal dangerous?

貧しい【まずしい】 poor 貧乏【びんぼう】な poor What's the difference?

Is it legal for a bar bouncer to confiscate a fake ID

GroupBy operation using an entire dataframe to group values

How can I get an unreasonable manager to approve time off?

What to do when surprise and a high initiative roll conflict with the narrative?

How to hide rifle during medieval town entrance inspection?

Traversing Oceania: A Cryptic Journey

Determining fair price for profitable mobile app business

Is White controlling this game?

Is it possible for a vehicle to be manufactured without a catalytic converter?

Are there any important biographies of nobodies?

How did old MS-DOS games utilize various graphic cards?

What ways have you found to get edits from non-LaTeX users?

Heap allocation on microcontroller

Why can I traceroute to this IP address, but not ping?

What is the color of artificial intelligence?

Should I ask for an extra raise?

How to make Keras-GAN work with non-standard input shape?

Why do Keras Conv1D layers' output tensors not have the input dimension?Maxpooling Layer causes error in KerasKeras layer shape in plot_model()Output of conv2d in kerasFine-tuning VGG, got:Negative dimension size caused by subtracting 2 from 1Why keras embeding layer has no keysThe output NN is image an image with values 0 or 1, but the expected are a range of integers between 0 and 255understanding output shape of keras Conv2DTransposepadding based on shape of largest layer of Concatenate inputs?How can I freeze last layer of my own model?

.everyoneloves__top-leaderboard:empty,.everyoneloves__mid-leaderboard:empty,.everyoneloves__bot-mid-leaderboard:empty height:90px;width:728px;box-sizing:border-box;

0

I have been using this Keras-GAN repository, specifically its pix2pix package for image-to-image translation. It works great when the input shape is a standard (256,256), and also works with the (28,28) MNIST data. However, when I try with inputs of shape (196,208), I get the following ValueError:

A 'Concatenate' layer requires inputs with matching shapes except for the concat axis. Got inputs shapes: [(None, 8, 8, 512), (None, 7, 7, 512)]

For some context, the Concatenate function is being called in the following block of code:

# Image input
d0 = Input(shape=self.img_shape)

# Downsampling
d1 = conv2d(d0, self.gf, bn=False)
d2 = conv2d(d1, self.gf*2)
d3 = conv2d(d2, self.gf*4)
d4 = conv2d(d3, self.gf*8)
d5 = conv2d(d4, self.gf*8)
d6 = conv2d(d5, self.gf*8)
d7 = conv2d(d6, self.gf*8)

# Upsampling
u1 = deconv2d(d7, d6, self.gf*8)
u2 = deconv2d(u1, d5, self.gf*8)
u3 = deconv2d(u2, d4, self.gf*8)
u4 = deconv2d(u3, d3, self.gf*4)
u5 = deconv2d(u4, d2, self.gf*2)
u6 = deconv2d(u5, d1, self.gf)

where conv2d and deconv2d are the following:

def conv2d(layer_input, filters, f_size=4, bn=True):
 """Layers used during downsampling"""
 d = Conv2D(filters, kernel_size=f_size, strides=2, padding='same')(layer_input)
 d = LeakyReLU(alpha=0.2)(d)
 if bn:
 d = BatchNormalization(momentum=0.8)(d)
 return d

def deconv2d(layer_input, skip_input, filters, f_size=4, dropout_rate=0):
 """Layers used during upsampling"""
 u = UpSampling2D(size=2)(layer_input)
 u = Conv2D(filters, kernel_size=f_size, strides=1, padding='same', activation='relu')(u)
 if dropout_rate:
 u = Dropout(dropout_rate)(u)
 u = BatchNormalization(momentum=0.8)(u)
 u = Concatenate()([u, skip_input])
 return u

keras

share|improve this question

asked Mar 24 at 19:00

Nick LNick L

11

add a comment | 

0

I have been using this Keras-GAN repository, specifically its pix2pix package for image-to-image translation. It works great when the input shape is a standard (256,256), and also works with the (28,28) MNIST data. However, when I try with inputs of shape (196,208), I get the following ValueError:

A 'Concatenate' layer requires inputs with matching shapes except for the concat axis. Got inputs shapes: [(None, 8, 8, 512), (None, 7, 7, 512)]

For some context, the Concatenate function is being called in the following block of code:

# Image input
d0 = Input(shape=self.img_shape)

# Downsampling
d1 = conv2d(d0, self.gf, bn=False)
d2 = conv2d(d1, self.gf*2)
d3 = conv2d(d2, self.gf*4)
d4 = conv2d(d3, self.gf*8)
d5 = conv2d(d4, self.gf*8)
d6 = conv2d(d5, self.gf*8)
d7 = conv2d(d6, self.gf*8)

# Upsampling
u1 = deconv2d(d7, d6, self.gf*8)
u2 = deconv2d(u1, d5, self.gf*8)
u3 = deconv2d(u2, d4, self.gf*8)
u4 = deconv2d(u3, d3, self.gf*4)
u5 = deconv2d(u4, d2, self.gf*2)
u6 = deconv2d(u5, d1, self.gf)

where conv2d and deconv2d are the following:

def conv2d(layer_input, filters, f_size=4, bn=True):
 """Layers used during downsampling"""
 d = Conv2D(filters, kernel_size=f_size, strides=2, padding='same')(layer_input)
 d = LeakyReLU(alpha=0.2)(d)
 if bn:
 d = BatchNormalization(momentum=0.8)(d)
 return d

def deconv2d(layer_input, skip_input, filters, f_size=4, dropout_rate=0):
 """Layers used during upsampling"""
 u = UpSampling2D(size=2)(layer_input)
 u = Conv2D(filters, kernel_size=f_size, strides=1, padding='same', activation='relu')(u)
 if dropout_rate:
 u = Dropout(dropout_rate)(u)
 u = BatchNormalization(momentum=0.8)(u)
 u = Concatenate()([u, skip_input])
 return u

keras

share|improve this question

asked Mar 24 at 19:00

Nick LNick L

11

add a comment | 

I have been using this Keras-GAN repository, specifically its pix2pix package for image-to-image translation. It works great when the input shape is a standard (256,256), and also works with the (28,28) MNIST data. However, when I try with inputs of shape (196,208), I get the following ValueError:

A 'Concatenate' layer requires inputs with matching shapes except for the concat axis. Got inputs shapes: [(None, 8, 8, 512), (None, 7, 7, 512)]

For some context, the Concatenate function is being called in the following block of code:

# Image input
d0 = Input(shape=self.img_shape)

# Downsampling
d1 = conv2d(d0, self.gf, bn=False)
d2 = conv2d(d1, self.gf*2)
d3 = conv2d(d2, self.gf*4)
d4 = conv2d(d3, self.gf*8)
d5 = conv2d(d4, self.gf*8)
d6 = conv2d(d5, self.gf*8)
d7 = conv2d(d6, self.gf*8)

# Upsampling
u1 = deconv2d(d7, d6, self.gf*8)
u2 = deconv2d(u1, d5, self.gf*8)
u3 = deconv2d(u2, d4, self.gf*8)
u4 = deconv2d(u3, d3, self.gf*4)
u5 = deconv2d(u4, d2, self.gf*2)
u6 = deconv2d(u5, d1, self.gf)

where conv2d and deconv2d are the following:

def conv2d(layer_input, filters, f_size=4, bn=True):
 """Layers used during downsampling"""
 d = Conv2D(filters, kernel_size=f_size, strides=2, padding='same')(layer_input)
 d = LeakyReLU(alpha=0.2)(d)
 if bn:
 d = BatchNormalization(momentum=0.8)(d)
 return d

def deconv2d(layer_input, skip_input, filters, f_size=4, dropout_rate=0):
 """Layers used during upsampling"""
 u = UpSampling2D(size=2)(layer_input)
 u = Conv2D(filters, kernel_size=f_size, strides=1, padding='same', activation='relu')(u)
 if dropout_rate:
 u = Dropout(dropout_rate)(u)
 u = BatchNormalization(momentum=0.8)(u)
 u = Concatenate()([u, skip_input])
 return u

keras

share|improve this question

asked Mar 24 at 19:00

Nick LNick L

11

# Image input
d0 = Input(shape=self.img_shape)

# Downsampling
d1 = conv2d(d0, self.gf, bn=False)
d2 = conv2d(d1, self.gf*2)
d3 = conv2d(d2, self.gf*4)
d4 = conv2d(d3, self.gf*8)
d5 = conv2d(d4, self.gf*8)
d6 = conv2d(d5, self.gf*8)
d7 = conv2d(d6, self.gf*8)

# Upsampling
u1 = deconv2d(d7, d6, self.gf*8)
u2 = deconv2d(u1, d5, self.gf*8)
u3 = deconv2d(u2, d4, self.gf*8)
u4 = deconv2d(u3, d3, self.gf*4)
u5 = deconv2d(u4, d2, self.gf*2)
u6 = deconv2d(u5, d1, self.gf)

def conv2d(layer_input, filters, f_size=4, bn=True):
 """Layers used during downsampling"""
 d = Conv2D(filters, kernel_size=f_size, strides=2, padding='same')(layer_input)
 d = LeakyReLU(alpha=0.2)(d)
 if bn:
 d = BatchNormalization(momentum=0.8)(d)
 return d

def deconv2d(layer_input, skip_input, filters, f_size=4, dropout_rate=0):
 """Layers used during upsampling"""
 u = UpSampling2D(size=2)(layer_input)
 u = Conv2D(filters, kernel_size=f_size, strides=1, padding='same', activation='relu')(u)
 if dropout_rate:
 u = Dropout(dropout_rate)(u)
 u = BatchNormalization(momentum=0.8)(u)
 u = Concatenate()([u, skip_input])
 return u

share|improve this question

asked Mar 24 at 19:00

Nick LNick L

11

share|improve this question

asked Mar 24 at 19:00

Nick LNick L

11

asked Mar 24 at 19:00

Nick LNick L

11

asked Mar 24 at 19:00

Nick LNick L

11