model.py

# -*- coding: utf-8 -*-
"""model.ipynb

Automatically generated by Colaboratory.

Original file is located at
    https://colab.research.google.com/drive/12h-cz3T6vOnmygh_17p1dmWSxp0H9Lqn
"""

import torch.nn as nn
import torch.nn.functional as F

# ********************************************S7 MODEL***************************************************************
class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()

        self.convblock1 = nn.Sequential(
            nn.Conv2d(in_channels=3, out_channels=16, kernel_size=(3, 3), padding=1, bias=False), #Rf = 3, j = 1
            nn.ReLU(),
            nn.BatchNorm2d(16),
            nn.Dropout(p = 0.1),

            nn.Conv2d(in_channels=16, out_channels=32, kernel_size=(3, 3), padding=1, bias=False), #Rf = 5, j = 1
            nn.ReLU(),
            nn.BatchNorm2d(32),
            nn.Dropout(p = 0.1),

           
            
            )
        self.convblock2 = nn.Sequential(

            #Dilated Network
           nn.Conv2d(in_channels=32, out_channels=16, kernel_size=(3, 3), padding=1,dilation=2, bias=False), #jout=2, kernel_size = 5, rf = 6+(4)*2 = 14, o/p = 14
            nn.ReLU(),
            nn.BatchNorm2d(16),
            nn.Dropout(p = 0.1),

            nn.Conv2d(in_channels=16, out_channels=32, kernel_size=(3, 3), padding=1, bias=False), #jout=2, kernel_size = 3, rf = 14+(2)*2 = 18, o/p =14
            nn.ReLU(),
            nn.BatchNorm2d(32),
            nn.Dropout(p = 0.1),

            
            )

        self.convblock3 = nn.Sequential(
            #DepthWise Seperable Network
            nn.Conv2d(in_channels=32, out_channels=64, kernel_size=(3, 3), padding=1, bias=False, groups = 32), # jout = 4, rf = 20+(2)*4 = 28, o/p = 7
            nn.ReLU(),
            nn.BatchNorm2d(64),
            nn.Dropout(p = 0.1),

            nn.Conv2d(in_channels=64, out_channels=64, kernel_size=(1, 1), padding=1, bias=False), #jout = 4, rf = 28+(0)*4 = 28, o/p = 9
            nn.ReLU(),
            nn.BatchNorm2d(64),
            nn.Dropout(p = 0.1),

            #DepthWise Seperable Network

            nn.Conv2d(in_channels=64, out_channels=128, kernel_size=(3, 3), padding=1,groups=64, bias=False),#jout = 4, rf = 28+(2)*4 = 36, o/p = 9
            nn.ReLU(),
            nn.BatchNorm2d(128),
            nn.Dropout(p = 0.1),

            nn.Conv2d(in_channels=128, out_channels=128, kernel_size=(1, 1), padding=1, bias=False), #jin = jout = 4, rf = 36+(0)*4 = 36, o/p = 11
            nn.ReLU(),
            nn.BatchNorm2d(128),
            nn.Dropout(p = 0.1),
            
            )

        self.convblock4 = nn.Sequential(

            #Dialated Network
          nn.Conv2d(in_channels=128, out_channels=64, kernel_size=(3, 3), padding=1,dilation=2, bias=False), #jout = 8, rf = 40+(4)*8 = 72, o/p = 3
            nn.ReLU(),
            nn.BatchNorm2d(64),
            nn.Dropout(p=0.15),


            #AVG Pool

            nn.AdaptiveAvgPool2d(1), #op = 1



           nn.Conv2d(in_channels=64, out_channels=10, kernel_size=(1, 1), padding=0, bias=False)#Op_size = 1, 
            

            
            )



        self.pool = nn.MaxPool2d(2, 2)

      


    def forward(self, x):

        x = self.convblock1(x) # i/p= 32 o/p=32 Rf = 6
        x = self.pool(x) # jout = 2, Rf = 6, O/p = 16 
        x = self.convblock2(x) # Rf = 18 jout = 2, o/p =14
        x = self.pool(x) # jout = 4, s = 2, Rf = 18+1*2 = 20, o/p = 7
        x = self.convblock3(x) # jout = 4, Rf = 36, o/p = 11
        x = self.pool(x) # jout = 8, s = 2, Rf = 36+1*4 = 40 o/p = 5
        x = self.convblock4(x)  # o/p = 1
     
        x = x.view(-1, 10)
      
        return x


#******************************************RESNET*******************************************************************

class BasicBlock(nn.Module):
    expansion = 1
 
    def __init__(self, in_planes, planes, stride=1, dropout=0.0):
        super(BasicBlock, self).__init__()
        self.conv1 = nn.Conv2d(in_planes, planes, kernel_size=3, stride=stride, padding=1, bias=False)
        self.bn1 = nn.BatchNorm2d(planes)
        self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=1, padding=1, bias=False)
        self.bn2 = nn.BatchNorm2d(planes)
 
        self.shortcut = nn.Sequential()
        if stride != 1 or in_planes != self.expansion*planes:
            self.shortcut = nn.Sequential(
                nn.Conv2d(in_planes, self.expansion*planes, kernel_size=1, stride=stride, bias=False),
                nn.BatchNorm2d(self.expansion*planes)
            )
        self.dropout = dropout
 
    def forward(self, x):
        out = F.relu(self.bn1(self.conv1(x)))
        out = F.dropout(out, p=self.dropout)
        out = self.bn2(self.conv2(out))
        out = F.dropout(out, p=self.dropout)
        out += self.shortcut(x)
        out = F.relu(out)
        out = F.dropout(out, p=self.dropout)
        return out
 
 
class Bottleneck(nn.Module):
    expansion = 4
 
    def __init__(self, in_planes, planes, stride=1):
        super(Bottleneck, self).__init__()
        self.conv1 = nn.Conv2d(in_planes, planes, kernel_size=1, bias=False)
        self.bn1 = nn.BatchNorm2d(planes)
        self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=stride, padding=1, bias=False)
        self.bn2 = nn.BatchNorm2d(planes)
        self.conv3 = nn.Conv2d(planes, self.expansion*planes, kernel_size=1, bias=False)
        self.bn3 = nn.BatchNorm2d(self.expansion*planes)
 
        self.shortcut = nn.Sequential()
        if stride != 1 or in_planes != self.expansion*planes:
            self.shortcut = nn.Sequential(
                nn.Conv2d(in_planes, self.expansion*planes, kernel_size=1, stride=stride, bias=False),
                nn.BatchNorm2d(self.expansion*planes)
            )
 
    def forward(self, x):
        out = F.relu(self.bn1(self.conv1(x)))
        out = F.relu(self.bn2(self.conv2(out)))
        out = self.bn3(self.conv3(out))
        out += self.shortcut(x)
        out = F.relu(out)
        return out
 
 
class ResNet(nn.Module):
    def __init__(self, block, num_blocks, num_classes=10, dropout=0.0):
        super(ResNet, self).__init__()
        self.in_planes = 64
        self.dropout = dropout
 
        self.conv1 = nn.Conv2d(3, 64, kernel_size=3, stride=1, padding=1, bias=False)
        self.bn1 = nn.BatchNorm2d(64)
        self.layer1 = self._make_layer(block, 64, num_blocks[0], stride=1)
        self.layer2 = self._make_layer(block, 128, num_blocks[1], stride=2)
        self.layer3 = self._make_layer(block, 256, num_blocks[2], stride=2)
        self.layer4 = self._make_layer(block, 512, num_blocks[3], stride=2)
        self.linear = nn.Linear(512*block.expansion, num_classes)
 
    def _make_layer(self, block, planes, num_blocks, stride):
        strides = [stride] + [1]*(num_blocks-1)
        layers = []
        for stride in strides:
            layers.append(block(self.in_planes, planes, stride, dropout=self.dropout))
            self.in_planes = planes * block.expansion
        return nn.Sequential(*layers)
 
    def forward(self, x):
        out = F.relu(self.bn1(self.conv1(x)))
        out = F.dropout(out, p=self.dropout)
        out = self.layer1(out)
        out = self.layer2(out)
        out = self.layer3(out)
        out = self.layer4(out)
        out = F.avg_pool2d(out, 4)
        out = out.view(out.size(0), -1)
        out = self.linear(out)
        return out
 
 
def ResNet18(dropout=0.15):
    return ResNet(BasicBlock, [2,2,2,2], dropout=dropout)