block(self.inplanes, planes, track_running_stats=True))
return nn.Sequential(*layers)
def run(self, x, target=None):
"""Function for getting the outputs of intermediate layers
"""
if target is None or target > 5:
raise NotImplementedError(
'Target is expected to be smaller than 6')
layers = [
self.layer1, self.layer2, self.layer3, self.layer4, self.layer5
]
for layer in layers[:target]:
x = layer(x)
return x
def forward(self, x):
if self.sobel is not None:
x = self.sobel(x)
x = self.layer1(x)
x = self.layer2(x)
x = self.layer3(x)
x = self.layer4(x)
x = self.layer5(x)
x = self.avgpool(x)
return map(lambda head: head(x), self.heads)
register('resnet34', ResNet34)
self.inplanes = planes * block.expansion
for i in range(1, blocks):
layers.append(
block(self.inplanes, planes, track_running_stats=True))
return nn.Sequential(*layers)
def forward(self, x, hidx, output_features=False):
if self.sobel is not None:
x = self.sobel(x)
elif self.grayscale is not None:
x = self.grayscale(x)
if output_features:
layers = [
self.layer1, self.layer2, self.layer3, self.layer4,
self.layer5, self.avgpool
]
else:
layers = [
self.layer1, self.layer2, self.layer3, self.layer4,
self.layer5, self.avgpool, self.heads[hidx]
]
for layer in layers:
x = layer(x)
return x
register('resnet34standard', ResNet34Standard)
layers.append(block(self.inplanes, planes, stride, downsample))
self.inplanes = planes * block.expansion
for i in range(1, blocks):
layers.append(block(self.inplanes, planes))
return nn.Sequential(*layers)
def forward(self, x, hidx, output_features=False):
if self.sobel is not None:
x = self.sobel(x)
elif self.grayscale is not None:
x = self.grayscale(x)
if output_features:
layers = [
self.layer1, self.layer2, self.layer3, self.layer4,
self.layer5, self.avgpool
]
else:
layers = [
self.layer1, self.layer2, self.layer3, self.layer4,
self.layer5, self.avgpool, self.heads[hidx]
]
for layer in layers:
x = layer(x)
return x
register('resnet50small', ResNet50Small)
layers.append(block(self.inplanes, planes, stride, downsample))
self.inplanes = planes * block.expansion
for i in range(1, blocks):
layers.append(block(self.inplanes, planes))
return nn.Sequential(*layers)
def forward(self, x, hidx, output_features=False):
if self.sobel is not None:
x = self.sobel(x)
elif self.grayscale is not None:
x = self.grayscale(x)
if output_features:
layers = [
self.layer1, self.layer2, self.layer3, self.layer4,
self.layer5, self.avgpool
]
else:
layers = [
self.layer1, self.layer2, self.layer3, self.layer4,
self.layer5, self.avgpool, self.heads[hidx]
]
for layer in layers:
x = layer(x)
return x
register('resnet50standard', ResNet50Standard)
layers.append(block(self.inplanes, planes, stride, downsample))
self.inplanes = planes * block.expansion
for i in range(1, blocks):
layers.append(block(self.inplanes, planes))
return nn.Sequential(*layers)
def forward(self, x, hidx, output_features=False):
if self.sobel is not None:
x = self.sobel(x)
elif self.grayscale is not None:
x = self.grayscale(x)
if output_features:
layers = [
self.layer1, self.layer2, self.layer3, self.layer4,
self.layer5, self.layer6, self.avgpool
]
else:
layers = [
self.layer1, self.layer2, self.layer3, self.layer4,
self.layer5, self.layer6, self.avgpool, self.heads[hidx]
]
for layer in layers:
x = layer(x)
return x
register('resnet50large', ResNet50Large)
downsample = nn.Sequential(
nn.Conv2d(self.inplanes, planes * block.expansion, kernel_size=1, stride=stride, bias=False),
nn.BatchNorm2d(planes * block.expansion, track_running_stats=True))
layers = []
layers.append(block(self.inplanes, planes, stride, downsample, track_running_stats=True))
self.inplanes = planes * block.expansion
for i in range(1, blocks):
layers.append(block(self.inplanes, planes, track_running_stats=True))
return nn.Sequential(*layers)
def forward(self, x, hidx, output_features=False):
if self.sobel is not None:
x = self.sobel(x)
elif self.grayscale is not None:
x = self.grayscale(x)
if output_features:
layers = [self.layer1, self.layer2, self.layer3, self.layer4, self.layer5, self.layer6, self.avgpool]
else:
layers = [self.layer1, self.layer2, self.layer3, self.layer4, self.layer5, self.layer6, self.avgpool,
self.heads[hidx]]
for layer in layers:
x = layer(x)
return x
register('resnet34large', ResNet34Large)