def __init__(self,
backbone='resnet',
output_stride=16,
num_classes=21,
nInputChannels=3,
sync_bn=True,
freeze_bn=False):
super(SegmentationNetwork, self).__init__()
output_shape = 128
channel_settings = [512, 1024, 512, 256]
self.Coarse_net = CoarseNet(channel_settings, output_shape,
num_classes)
self.Fine_net = FineNet(channel_settings[-1], output_shape,
num_classes)
BatchNorm = nn.BatchNorm2d
self.backbone = build_backbone(backbone,
output_stride,
BatchNorm,
nInputChannels,
pretrained=False)
self.psp4 = PSPModule(in_features=2048,
out_features=512,
sizes=(1, 2, 3, 6),
n_classes=256)
self.upsample = nn.Upsample(size=(512, 512),
mode='bilinear',
align_corners=True)
if freeze_bn:
self.freeze_bn()
def __init__(self, backbone='resnet_edge', output_stride=16, num_classes=21,
sync_bn=True, freeze_bn=False):
super(DeepLab, self).__init__()
if backbone == 'drn':
output_stride = 8
if freeze_bn ==True:
print("Use frozen BN in DeepLab")
BatchNorm=FrozenBatchNorm2d
elif sync_bn == True:
BatchNorm = SynchronizedBatchNorm2d
else:
BatchNorm = nn.BatchNorm2d
self.backbone = build_backbone(backbone, output_stride, BatchNorm)
self.aspp = build_aspp(backbone, output_stride, BatchNorm)
self.decoder = build_decoder(num_classes, backbone, BatchNorm)
def __init__(self, backbone='resnet', output_stride=16, num_classes=21,
sync_bn=False, freeze_bn=False):
super(DeepLab, self).__init__()
if backbone == 'drn':
output_stride = 8
if sync_bn:
BatchNorm = SynchronizedBatchNorm2d
else:
BatchNorm = nn.BatchNorm2d
self.backbone = build_backbone(backbone, output_stride, BatchNorm)
self.aspp = build_aspp(backbone, output_stride, BatchNorm)
self.decoder = build_decoder(num_classes, backbone, BatchNorm)
if freeze_bn:
self.freeze_bn()
def __init__(self, backbone='resnet', output_stride=16, num_classes=21,
sync_bn=False, num_domain=3, freeze_bn=False, lam =0.9):
super(DeepLab, self).__init__()
if backbone == 'drn':
output_stride = 8
if sync_bn == True:
BatchNorm = SynchronizedBatchNorm2d
else:
BatchNorm = nn.BatchNorm2d
self.lam = lam
self.centroids = nn.Parameter(torch.randn(num_domain, 304, 64, 64), requires_grad=False)
self.backbone = build_backbone(backbone, output_stride, BatchNorm)
self.aspp = build_aspp(backbone, output_stride, BatchNorm)
self.decoder = build_decoder(num_classes, num_domain, backbone, BatchNorm)
self.last_conv_mask = nn.Sequential(BatchNorm(3),
nn.ReLU(),
nn.Dropout(0.5),
nn.Conv2d(3, num_domain, kernel_size=1, stride=1))
# build encoder for domain code
self.encoder_d = build_encoderDC(num_domain, backbone, BatchNorm)
if freeze_bn:
self.freeze_bn()
def __init__(self,
backbone='resnet',
output_stride=16,
num_classes=21,
sync_bn=True,
freeze_bn=False,
dcn=False,
g_in_dim=3):
super(DeepLab, self).__init__()
if sync_bn == True:
BatchNorm = SynchronizedBatchNorm2d
else:
BatchNorm = nn.BatchNorm2d
self.backbone = build_backbone(backbone,
output_stride,
BatchNorm,
g_in_dim=g_in_dim)
self.aspp = build_aspp(backbone, output_stride, BatchNorm, dcn)
self.decoder = build_decoder(num_classes, backbone, BatchNorm, dcn)
self.freeze_bn = freeze_bn