def resnet50(pretrained=False, progress=True, **kwargs):
"""Constructs a ResNet-50 model.
Args:
pretrained (bool): If True, returns a model pre-trained on ImageNet
progress (bool): If True, displays a progress bar of the download to stderr
"""
model = _resnet('resnet50', Bottleneck, [3, 4, 6, 3], progress=progress, pretrained=False, **kwargs)
if pretrained:
model = load_and_modify_pretrained_num_classes(model, model_urls['resnet50'], kwargs['num_classes'])
return model
def seatt_resnet50(pretrained=True, **kwargs):
model = SeAtt(SEResNetBottleneck,
layers=[3, 4, 6, 3],
att_layers=[1, 2, 5],
groups=1,
reduction=16,
dropout_p=None,
inplanes=64,
input_3x3=False,
downsample_kernel_size=1,
downsample_padding=0)
if pretrained:
model = load_and_modify_pretrained_num_classes(
model, model_urls['seatt_resnet50'], kwargs['num_classes'])
return model
def seatt154(pretrained=True, **kwargs):
model = SeAtt(block=SEBottleneck,
layers=[3, 8, 36, 3],
att_layers=[1, 2, 5],
groups=64,
reduction=16,
dropout_p=0.2,
inplanes=128,
input_3x3=True,
downsample_kernel_size=3,
downsample_padding=1)
if pretrained:
model = load_and_modify_pretrained_num_classes(model,
model_urls['se154'],
kwargs['num_classes'])
return model
def seatt_resnext50_base(pretrained=True, **kwargs):
"""
The difference with `seatt_resnext50_32x4d` is the number of attention module layer
"""
model = SeAtt(block=SEResNeXtBottleneck,
layers=[3, 4, 6, 3],
att_layers=[1, 1, 1],
groups=32,
reduction=16,
dropout_p=None,
inplanes=64,
input_3x3=False,
downsample_kernel_size=1,
downsample_padding=0)
if pretrained:
model = load_and_modify_pretrained_num_classes(
model, model_urls['se_resnext50_32x4d'], kwargs['num_classes'])
return model
def resnext50_32x4d(pretrained=True, progress=True, **kwargs):
model = _resnet('resnext50_32x4d', Bottleneck, [3, 4, 6, 3], groups=32, width_per_group=4,
pretrained=pretrained, progress=progress)
if pretrained:
model = load_and_modify_pretrained_num_classes(model, model_urls['resnext50_32x4d'], kwargs['num_classes'])
return model