CNN

skin-cancer-recognition.py

class CNN(nn.Module):
    def __init__(self, num_classes):
        super(CNN, self).__init__()
        
        self.CONV1 = 64
        self.CONV2 = 128
        self.CONV3 = 256
        self.CONV4 = 512
        self.FC1   = 64
        
        self.conv_layers = nn.Sequential(
            nn.Conv2d(3, self.CONV1, kernel_size=3, padding=1),
            nn.BatchNorm2d(self.CONV1),
            nn.LeakyReLU(0.1),
            nn.MaxPool2d(2, 2),

            nn.Conv2d(self.CONV1, self.CONV2, kernel_size=3, padding=1),
            nn.BatchNorm2d(self.CONV2),
            nn.LeakyReLU(0.1),
            nn.MaxPool2d(2, 2),

            nn.Conv2d(self.CONV2, self.CONV3, kernel_size=3, padding=1),
            nn.BatchNorm2d(self.CONV3),
            nn.LeakyReLU(0.1),
            nn.MaxPool2d(2, 2),

            nn.Conv2d(self.CONV3, self.CONV4, kernel_size=3, padding=1),
            nn.BatchNorm2d(self.CONV4),
            nn.LeakyReLU(0.1),
            nn.MaxPool2d(2, 2),
        )
        
        self.gap = nn.AdaptiveAvgPool2d(1)
        self.fc1 = nn.Linear(self.CONV4, self.FC1)
        self.dropout1 = nn.Dropout(p=0.2)
        self.fc2 = nn.Linear(self.FC1, num_classes)
        self.dropout2 = nn.Dropout(p=0.1)
        
    def forward(self, x):
        x = self.conv_layers(x)
        x = self.gap(x)
        x = x.view(x.size(0), -1)
        x = F.leaky_relu(self.fc1(x), 0.1)
        x = self.dropout1(x)
        x = self.fc2(x)
        x = self.dropout2(x)
        return x

    def predict(self, x):
        logits = self.forward(x)
        return F.softmax(logits, dim=1)