def __init__(self, num_classes=8, pretrained=False):
super(AlexNet, self).__init__()
self.num_classes = num_classes
self.alexnet = alexnet.AlexNet(num_classes=num_classes)
self.dim = self.alexnet.dim
if pretrained:
utl.load_state_dict(self.alexnet.state_dict(),
model_zoo.load_url(model_urls['alexnet']))
nn.init.xavier_normal(self.alexnet.classifier[6].weight)
def inception_v4(pretrained=False, **kwargs):
if pretrained:
if 'transform_input' not in kwargs:
kwargs['transform_input'] = True
model = InceptionV4(**kwargs)
utl.load_state_dict(
model.state_dict(),
model_zoo.load_url(model_urls['inception_v4_google']))
return model
return InceptionV4(**kwargs)
def nasnetalarge(pretrained=False, **kwargs):
r"""NASNetALarge model architecture from the
`"NASNet" <https://arxiv.org/abs/1707.07012>`_ paper.
"""
if pretrained:
model = NASNetALarge(**kwargs)
utl.load_state_dict(model.state_dict(),
model_zoo.load_url(model_urls['nasnetalarge']))
return model
return NASNetALarge(**kwargs)
def __init__(self, num_classes=8, pretrained=False):
super(ResNet18, self).__init__()
self.num_classes = num_classes
blocbasic = resnet.BasicBlock
cfg = [2, 2, 2, 2]
self.resnet18 = resnet.ResNet(blocbasic, cfg, num_classes=num_classes)
self.dim = self.resnet18.dim
if pretrained:
utl.load_state_dict(self.resnet18.state_dict(),
model_zoo.load_url(model_urls['resnet18']))
nn.init.xavier_normal(self.resnet18.fc.weight)
def __init__(self, num_classes=8, pretrained=False):
super(VGG11, self).__init__()
self.num_classes = num_classes
cfg = [64, 'M', 128, 'M', 256, 256, 'M', 512, 512, 'M', 512, 512, 'M']
features = vgg.make_layers(cfg, batch_norm=False)
self.vgg11 = vgg.VGG(features, num_classes=num_classes)
self.dim = self.vgg11.dim
if pretrained:
utl.load_state_dict(self.vgg11.state_dict(),
model_zoo.load_url(model_urls['vgg11']))
nn.init.xavier_normal(self.vgg11.classifier[6].weight)
def resnet101(pretrained=False, **kwargs):
"""Constructs a ResNet-101 model.
Args:
pretrained (bool): If True, returns a model pre-trained on ImageNet
"""
model = ResNet(Bottleneck, [3, 4, 23, 3], **kwargs)
if pretrained:
#model.load_state_dict(model_zoo.load_url(model_urls['resnet101']))
utl.load_state_dict(model.state_dict(),
model_zoo.load_url(model_urls['resnet101']))
return model
def resnet34(pretrained=False, **kwargs):
"""Constructs a ResNet-34 model.
Args:
pretrained (bool): If True, returns a model pre-trained on ImageNet
"""
model = ResNet(BasicBlock, [3, 4, 6, 3], **kwargs)
if pretrained:
#model.load_state_dict(model_zoo.load_url(model_urls['resnet34']))
utl.load_state_dict(model.state_dict(),
model_zoo.load_url(model_urls['resnet34']))
return model
def __init__(self, num_classes=8, pretrained=False):
super(InceptionV3, self).__init__()
self.num_classes = num_classes
self.inception = inception.Inception3(num_classes=num_classes,
transform_input=False,
aux_logits=False)
self.dim = self.inception.dim
if pretrained:
utl.load_state_dict(
self.inception.state_dict(),
model_zoo.load_url(model_urls['inception_v3_google']))
nn.init.xavier_normal(self.inception.fc.weight)
def __init__(self, num_classes=8, pretrained=False):
super(DenseNet, self).__init__()
self.num_classes = num_classes
self.densenet = torchvision.models.DenseNet(num_classes=num_classes,
num_init_features=64,
growth_rate=32,
block_config=(6, 12, 24,
16))
self.dim = self.densenet.dim
if pretrained:
utl.load_state_dict(self.densenet.state_dict(),
model_zoo.load_url(model_urls['densenet121']))
nn.init.xavier_normal(self.densenet.classifier.weight)
def vgg19_bn(pretrained=False, in_channels=3, **kwargs):
"""VGG 19-layer model (configuration 'E') with batch normalization
Args:
pretrained (bool): If True, returns a model pre-trained on ImageNet
"""
if pretrained:
kwargs['init_weights'] = False
model = VGG(make_layers(cfg['E'], in_channels, batch_norm=True), **kwargs)
if pretrained:
#model.load_state_dict(model_zoo.load_url(model_urls['vgg19_bn']))
utl.load_state_dict(model.state_dict(),
model_zoo.load_url(model_urls['vgg19_bn']))
return model
def vgg16(pretrained=False, in_channels=3, **kwargs):
"""VGG 16-layer model (configuration "D")
Args:
pretrained (bool): If True, returns a model pre-trained on ImageNet
"""
if pretrained:
kwargs['init_weights'] = False
model = VGG(make_layers(cfg['D'], in_channels), **kwargs)
if pretrained:
#model.load_state_dict(model_zoo.load_url(model_urls['vgg16']))
utl.load_state_dict(model.state_dict(),
model_zoo.load_url(model_urls['vgg16']))
return model
def densenet161(pretrained=False, **kwargs):
r"""Densenet-161 model from
`"Densely Connected Convolutional Networks" <https://arxiv.org/pdf/1608.06993.pdf>`_
Args:
pretrained (bool): If True, returns a model pre-trained on ImageNet
"""
model = DenseNet(num_init_features=96,
growth_rate=48,
block_config=(6, 12, 36, 24),
**kwargs)
if pretrained:
#model.load_state_dict(model_zoo.load_url(model_urls['densenet161']))
utl.load_state_dict(model.state_dict(),
model_zoo.load_url(model_urls['densenet161']))
return model
def inception_v3(pretrained=False, **kwargs):
r"""Inception v3 model architecture from
`"Rethinking the Inception Architecture for Computer Vision" <http://arxiv.org/abs/1512.00567>`_.
Args:
pretrained (bool): If True, returns a model pre-trained on ImageNet
"""
if pretrained:
if 'transform_input' not in kwargs:
kwargs['transform_input'] = True
model = Inception3(**kwargs)
#model.load_state_dict(model_zoo.load_url(model_urls['inception_v3_google']))
utl.load_state_dict(
model.state_dict(),
model_zoo.load_url(model_urls['inception_v3_google']))
return model
return Inception3(**kwargs)
