def __init__(self,
num_classes=2,
in_channels=3,
arch='resnet101',
output_stride=16,
bn_momentum=0.9,
freeze_bn=False,
pretrained=False,puzzle=3,
**kwargs):
super(DeepLab, self).__init__(**kwargs)
self.model_name = 'deeplabv3plus_' + arch
#num_classes=puzzle**2
# Setup arch
if arch == 'resnet18':
NotImplementedError('resnet18 backbone is not implemented yet.')
elif arch == 'resnet34':
NotImplementedError('resnet34 backbone is not implemented yet.')
elif arch == 'resnet50':
self.backbone = ResNet.resnet50(bn_momentum, pretrained)
if in_channels != 3:
self.backbone.conv1 = nn.Conv2d(
in_channels, 64, kernel_size=7, stride=2, padding=3, bias=False)
elif arch == 'resnet101':
self.backbone = ResNet.resnet101(bn_momentum, pretrained)
if in_channels != 3:
self.backbone.conv1 = nn.Conv2d(
in_channels, 64, kernel_size=7, stride=2, padding=3, bias=False)
self.encoder = Encoder(bn_momentum, output_stride)
#self.decoder = Decoder(num_classes, bn_momentum)
self.avgpool = nn.AdaptiveAvgPool2d((3, 3))
self.fc5 = nn.Sequential(
nn.Linear(256 * puzzle * puzzle, 256),
nn.ReLU(),
nn.Dropout(p=0.1)
)
self.fc6 = nn.Sequential(
nn.Linear(256 * (puzzle**2), 1024),
nn.ReLU(),
nn.Dropout(p=0.1)
)
self.classifier = nn.Sequential(
nn.Linear(1024, num_classes)
)
# projection head
'''
self.proj = nn.Sequential(
nn.Conv2d(256, 256, 1, bias=False),
nn.BatchNorm2d(256),
nn.ReLU(inplace=True),
nn.Conv2d(256, 10, 1, bias=True)
)
'''
self.proj =nn.Sequential(nn.Linear(256, 256), nn.ReLU(), nn.Linear(256,num_classes))
def __init__(self,
in_channels=3,
arch='resnet101',
output_stride=16,
bn_momentum=0.9,
freeze_bn=False,
pretrained=False,
**kwargs):
super(DeepLab, self).__init__(**kwargs)
self.model_name = 'deeplabv3plus_' + arch
# Setup arch
if arch == 'resnet18':
NotImplementedError('resnet18 backbone is not implemented yet.')
elif arch == 'resnet34':
NotImplementedError('resnet34 backbone is not implemented yet.')
elif arch == 'resnet50':
self.backbone = ResNet.resnet50(bn_momentum, pretrained)
if in_channels != 3:
self.backbone.conv1 = nn.Conv2d(in_channels,
64,
kernel_size=7,
stride=2,
padding=3,
bias=False)
elif arch == 'resnet101':
self.backbone = ResNet.resnet101(bn_momentum, pretrained)
if in_channels != 3:
self.backbone.conv1 = nn.Conv2d(in_channels,
64,
kernel_size=7,
stride=2,
padding=3,
bias=False)
self.encoder = Encoder(bn_momentum, output_stride)
# self.decoder = Decoder(num_classes, bn_momentum)
self.avgpool = nn.AdaptiveAvgPool2d((1, 1))
# projection head
'''
def __init__(self,
num_classes=2,
in_channels=3,
arch='resnet101',
output_stride=16,
bn_momentum=0.9,
freeze_bn=False,
pretrained=False,
puzzle=3,
**kwargs):
super(DeepLab, self).__init__(**kwargs)
self.model_name = 'deeplabv3plus_' + arch
#num_classes=puzzle**2
# Setup arch
if arch == 'resnet18':
NotImplementedError('resnet18 backbone is not implemented yet.')
elif arch == 'resnet34':
NotImplementedError('resnet34 backbone is not implemented yet.')
elif arch == 'resnet50':
self.backbone = ResNet.resnet50(bn_momentum, pretrained)
if in_channels != 3:
self.backbone.conv1 = nn.Conv2d(in_channels,
64,
kernel_size=7,
stride=2,
padding=3,
bias=False)
elif arch == 'resnet101':
self.backbone = ResNet.resnet101(bn_momentum, pretrained)
if in_channels != 3:
self.backbone.conv1 = nn.Conv2d(in_channels,
64,
kernel_size=7,
stride=2,
padding=3,
bias=False)
self.encoder = Encoder(bn_momentum, output_stride)
#self.decoder = Decoder(num_classes, bn_momentum)
self.decoder4 = nn.Sequential(
nn.ConvTranspose2d(256, 256, 5, 2, padding=2, output_padding=0),
nn.BatchNorm2d(256),
nn.ReLU(True),
)
self.decoder3 = nn.Sequential(
nn.ConvTranspose2d(256 + 1024,
256,
5,
2,
padding=2,
output_padding=1),
nn.BatchNorm2d(256),
nn.ReLU(True),
)
self.decoder2 = nn.Sequential(
nn.ConvTranspose2d(256 + 512,
256,
3,
2,
padding=1,
output_padding=1),
nn.BatchNorm2d(256),
nn.ReLU(True),
)
self.decoder1 = nn.Sequential(
nn.ConvTranspose2d(256, 64, 3, 2, padding=1, output_padding=1),
nn.BatchNorm2d(64),
nn.ReLU(True),
)
self.conv_final = nn.Sequential(
nn.Conv2d(64, in_channels, kernel_size=7, padding=3, bias=True),
nn.Tanh())
# projection head
'''
def __init__(self,
num_classes=2,
in_channels=3,
arch='resnet101',
output_stride=16,
bn_momentum=0.9,
freeze_bn=False,
pretrained=False,
patch_size=16,
patch_num=4,
patch_out_channel=False,
pross_num=28,
**kwargs):
super(DeepLab, self).__init__(**kwargs)
self.model_name = 'deeplabv3plus_' + arch
# Setup arch
if arch == 'resnet18':
NotImplementedError('resnet18 backbone is not implemented yet.')
elif arch == 'resnet34':
NotImplementedError('resnet34 backbone is not implemented yet.')
elif arch == 'resnet50':
self.backbone = ResNet.resnet50(bn_momentum, pretrained)
if in_channels != 3:
self.backbone.conv1 = nn.Conv2d(in_channels,
64,
kernel_size=7,
stride=2,
padding=3,
bias=False)
elif arch == 'resnet101':
self.backbone = ResNet.resnet101(bn_momentum, pretrained)
if in_channels != 3:
self.backbone.conv1 = nn.Conv2d(in_channels,
64,
kernel_size=7,
stride=2,
padding=3,
bias=False)
self.encoder = Encoder(bn_momentum, output_stride)
#self.decoder = Decoder(num_classes, bn_momentum)
self.avgpool = nn.AdaptiveAvgPool2d((1, 1))
self.patch_num = patch_num
self.patch_size = patch_size
self.pross_num = pross_num
# projection head
'''
self.proj = nn.Sequential(
nn.Conv2d(256, 256, 1, bias=False),
nn.BatchNorm2d(256),
nn.ReLU(inplace=True),
nn.Conv2d(256, 10, 1, bias=True)
)
'''
self.proj = nn.Sequential(nn.Linear(256 * 2, 256), nn.ReLU(),
nn.Linear(256, num_classes))
if not patch_out_channel:
patch_out_channel = num_classes
self.proj1 = nn.Sequential(nn.Linear(num_classes, num_classes),
nn.ReLU(),
nn.Linear(num_classes, patch_out_channel))