from torch import nn
[docs]class NetworkInNetwork(nn.Module):
"""
Implementation of a Network In Network model
:param num_classes (int): Number of classes for classification
:param in_channels (int): Number of channels in input specimens
"""
def __init__(self, num_classes=10, in_channels=3):
super(NetworkInNetwork, self).__init__()
self.num_classes = num_classes
self.in_channels = in_channels
self.features = nn.Sequential(
nn.Conv2d(self.in_channels, 192, 5, padding=2),
nn.ReLU(inplace=True),
nn.Conv2d(192, 160, 1),
nn.ReLU(inplace=True),
nn.Conv2d(160, 96, 1),
nn.ReLU(inplace=True),
nn.MaxPool2d(3, stride=2, ceil_mode=True),
nn.Dropout(inplace=True),
nn.Conv2d(96, 192, 5, padding=2),
nn.ReLU(inplace=True),
nn.Conv2d(192, 192, 1),
nn.ReLU(inplace=True),
nn.Conv2d(192, 192, 1),
nn.ReLU(inplace=True),
nn.AvgPool2d(3, stride=2, ceil_mode=True),
nn.Dropout(inplace=True),
nn.Conv2d(192, 192, 3, padding=1),
nn.ReLU(inplace=True),
nn.Conv2d(192, 192, 1),
nn.ReLU(inplace=True),
nn.Conv2d(192, self.num_classes, 1),
nn.ReLU(inplace=True),
nn.AvgPool2d(8, stride=1),
)
self._initialize_weights()
[docs] def forward(self, x):
x = self.features(x)
x = x.view(x.size(0), self.num_classes)
return x
def _initialize_weights(self):
for m in self.modules():
if isinstance(m, nn.Conv2d):
m.weight.data.normal_(0, 0.05)
if m.bias is not None:
m.bias.data.zero_()