from sconce.models.base import Model
from sconce.models.layers import FullyConnectedLayer
from torch import nn
from torch.nn import functional as F
[docs]class BasicAutoencoder(Model):
"""
A basic 2D image autoencoder built up of fully connected layers, three each in the encoder and the decoder.
Loss:
This model uses binary cross-entropy for the loss.
Metrics:
None
Arguments:
image_height (int): image height in pixels.
image_width (int): image width in pixels.
hidden_size (int): the number of activations in each of the 4 hidden layers.
latent_size (int): the number of activations in the latent representation (encoder output).
"""
def __init__(self, image_height, image_width, hidden_size, latent_size):
super().__init__()
self.num_pixels = image_height * image_width
self.fc1 = FullyConnectedLayer(in_size=self.num_pixels,
out_size=hidden_size,
activation=nn.ReLU())
self.fc2 = FullyConnectedLayer(in_size=hidden_size,
out_size=hidden_size,
activation=nn.ReLU())
self.fc3 = FullyConnectedLayer(in_size=hidden_size,
out_size=latent_size,
activation=nn.ReLU())
self.fc4 = FullyConnectedLayer(in_size=latent_size,
out_size=hidden_size,
activation=nn.ReLU())
self.fc5 = FullyConnectedLayer(in_size=hidden_size,
out_size=hidden_size,
activation=nn.ReLU())
self.fc6 = FullyConnectedLayer(in_size=hidden_size,
out_size=self.num_pixels,
activation=nn.Sigmoid())
[docs] def encode(self, inputs, **kwargs):
encoder_input = inputs.view(-1, self.num_pixels)
x = self.fc1(encoder_input)
x = self.fc2(x)
x_latent = self.fc3(x)
return x_latent
[docs] def decode(self, x_latent):
x = self.fc4(x_latent)
x = self.fc5(x)
outputs = self.fc6(x)
return outputs
[docs] def forward(self, inputs, **kwargs):
x_latent = self.encode(inputs)
outputs = self.decode(x_latent)
return {'outputs': outputs}
[docs] def calculate_loss(self, inputs, outputs, **kwargs):
reconstruction_loss = F.binary_cross_entropy(outputs,
inputs.view_as(outputs))
return {'loss': reconstruction_loss}