Class Heavy_Classifier

light_classifier.py:61–106

      return x

class Heavy_Classifier(torch.nn.Module):
   def __init__(self):
      super(Heavy_Classifier, self).__init__()
      self.conv1 = self.conv_block(c_in = 3, c_out = 25, kernel_size = 7, stride = 1, padding = 1)
      self.conv2 = self.conv_block(c_in = 25, c_out = 40, kernel_size = 5, stride = 1, padding = 1)
      self.conv3 = self.conv_block(c_in = 40, c_out = 40, kernel_size = 5, stride = 1, padding = 1)
      self.conv4 = self.conv_block(c_in = 40, c_out = 25, kernel_size = 3, stride = 1, padding = 1)
      self.conv5 = self.conv_block(c_in = 25, c_out = 25, kernel_size = 3, stride = 1, padding = 1)
      self.conv6 = self.conv_block(c_in = 25, c_out = 3, kernel_size = 3, stride = 1, padding = 1)
      # 28px --> ???
      # 32px --> 3
      # 100px --> 48 (heavy)
      self.bigN = 3
      self.fc1 = nn.Linear(self.bigN, 64)
      self.fc2 = nn.Linear(64, 9)
      self.fc3 = nn.Linear(9, 2)
      self.maxpool = nn.MaxPool2d(kernel_size=2, stride=2)

   def conv_block(self, c_in, c_out, dropout=0.1, kernel_size=3, stride=1, **kwargs):
      seq_block = nn.Sequential(
         nn.Conv2d(in_channels=c_in, out_channels=c_out, kernel_size=kernel_size, **kwargs),
         nn.BatchNorm2d(num_features=c_out),
         nn.ReLU(),
         nn.Dropout2d(p=dropout)
      )
      return seq_block

   def forward(self, x):
      x = self.conv1(x)
      x = self.maxpool(x)
      x = self.conv2(x)
      x = self.conv3(x)
      x = self.maxpool(x)
      x = self.conv4(x)
      x = self.conv5(x)
      x = self.maxpool(x)
      x = self.conv6(x)
      x = self.maxpool(x)
      x = x.reshape((-1, self.bigN))

      x = F.relu(self.fc1(x))
      x = F.relu(self.fc2(x))
      x = F.relu(self.fc3(x))
      # x = F.relu(x)
      # x = torch.sigmoid(self.fc3(x))
      return x

class Ultra_Light_Classifier(torch.nn.Module):
   def __init__(self):