Created
October 19, 2018 14:56
-
-
Save SrivastavaKshitij/e998bc90e0f0aa8875110cc87b047643 to your computer and use it in GitHub Desktop.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
import torch | |
from torch import nn | |
class autoencoder(nn.Module): | |
def __init__(self,downsizing_factor=None,in_channels=1): | |
self.downsize = downsizing_factor | |
self.in_channels = in_channels | |
super(autoencoder,self).__init__() | |
conv_modules=[] | |
self.in_channels = self.in_channels | |
self.out_channels = 4 * self.in_channels | |
self.block = [nn.Conv2d(self.in_channels,self.out_channels,3,stride=2,padding=1), | |
nn.ReLU(True), | |
nn.Conv2d(self.out_channels,int((self.out_channels)/2),1,stride=1),nn.ReLU(True)] | |
conv_modules.extend(self.block) | |
for i in range(1,self.downsize): | |
#print(f"in channels , out channel {self.in_channels,self.out_channels}") | |
self.in_channels=int((self.out_channels)/2) | |
self.out_channels = int(4 * self.in_channels) | |
self.block = [nn.Conv2d(self.in_channels,self.out_channels,3,stride=2,padding=1), | |
nn.ReLU(True), | |
nn.Conv2d(self.out_channels,int((self.out_channels)/2),1,stride=1),nn.ReLU(True)] | |
conv_modules.extend(self.block) | |
self.conv = nn.Sequential(*conv_modules) | |
## Doconv part | |
self.deconv_in_channels = int((self.out_channels)/2) | |
deconv_modules=[] | |
for i in range(self.downsize): | |
self.deconv_block = [nn.ConvTranspose2d(self.deconv_in_channels,int(self.deconv_in_channels/2),2,stride=2,padding=(0)), | |
nn.ReLU(True)] | |
deconv_modules.extend(self.deconv_block) | |
self.deconv_in_channels=int(self.deconv_in_channels/2) | |
self.deconv = nn.Sequential(*deconv_modules) | |
def forward(self,x): | |
x = self.conv(x) | |
return self.deconv(x) | |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
a = autoencoder(downsizing_factor=5,in_channels=1)
print(a)
b = a(torch.unsqueeze(torch.randn(1,512,512),0))
print(np.shape(b))