Adding nn.Dropout2d to each conv layer of a pretrained VGG model

vision
1

Hi all,

In my study, I am using pertained VGG models of Pytorch and use them for the classification of Imagenet dataset.

The thing is after each layer of conv layers of the VGG19bn model, I need to add a nn.Dropout2d layer. But I couldn’t figure out how to do it.

Can you please help me to enable dropout2d layer for pretrained vgg19bn model of Pytorch library?

I paste below code to express my need in a better way. After each conv2d, batchnorm2d, ReLU layers, I also want to add nn.Dropout2d() layer. I will be making some experiments about MC dropout sampling by enabling dropout during prediction time.

cfg = {
    'A': [64, 'M', 128, 'M', 256, 256, 'M', 512, 512, 'M', 512, 512, 'M'],
    'B': [64, 64, 'M', 128, 128, 'M', 256, 256, 'M', 512, 512, 'M', 512, 512, 'M'],
    'D': [64, 64, 'M', 128, 128, 'M', 256, 256, 256, 'M', 512, 512, 512, 'M', 512, 512, 512, 'M'],
    'E': [64, 64, 'M', 128, 128, 'M', 256, 256, 256, 256, 'M', 512, 512, 512, 512, 'M', 
          512, 512, 512, 512, 'M'],
}

def make_layers(cfg, batch_norm=False):
    layers = []
    in_channels = 3
    for v in cfg:
        if v == 'M':
            layers += [nn.MaxPool2d(kernel_size=2, stride=2)]
        else:
            conv2d = nn.Conv2d(in_channels, v, kernel_size=3, padding=1)
            if batch_norm:
                layers += [conv2d, nn.BatchNorm2d(v), nn.ReLU(inplace=True), nn.Dropout2d(p=0.1)]
                #layers += [conv2d, nn.BatchNorm2d(v), nn.ReLU(inplace=True)]
            else:
                layers += [conv2d, nn.ReLU(inplace=True), nn.Dropout2d(p=0.1)]
                #layers += [conv2d, nn.ReLU(inplace=True)]
            in_channels = v
    return nn.Sequential(*layers)

class VGG(nn.Module):
    '''
    VGG model 
    '''
    def __init__(self, features):
        super(VGG, self).__init__()
        self.features = features
        self.classifier = nn.Sequential(
            nn.Dropout(),
            nn.Linear(512, 512),
            nn.ReLU(True),
            nn.Dropout(),
            nn.Linear(512, 512),
            nn.ReLU(True),
            nn.Linear(512, 10),
        )
         # Initialize weights
        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
                m.weight.data.normal_(0, math.sqrt(2. / n))
                m.bias.data.zero_()


    def forward(self, x):
        x = self.features(x)
        x = x.view(x.size(0), -1)
        x = self.classifier(x)
        return x

def vgg19_bn():
    """VGG 19-layer model (configuration 'E') with batch normalization"""
    return VGG(make_layers(cfg['E'], batch_norm=True))

Using a pretrained model in Pytorch is as easy as below:

model_cnn = models.vgg19_bn(pretrained=True)

But after loading the pretrained model, how can I change it to enable dropout in conv. layers?

Any help is really appreciated as I need to solve this problem as soon as possible.

Regards…