def __init__(self,
backbone,
BatchNorm,
output_stride,
num_classes,
freeze_bn=False):
super(SplitDeepLabDANet, self).__init__()
self.backbone = backbone
self.aspp = build_aspp(backbone, output_stride, BatchNorm)
self.decoder1 = build_decoder(num_classes, backbone, BatchNorm)
self.decoder2 = build_decoder(num_classes, backbone, BatchNorm)
self.decoder3 = build_decoder(num_classes, backbone, BatchNorm)
self.decoder4 = build_decoder(num_classes, backbone, BatchNorm)
self.decoder5 = build_decoder(num_classes, backbone, BatchNorm)
self.decoders = [
self.decoder1, self.decoder2, self.decoder3, self.decoder4,
self.decoder5
]
self.output_stride = output_stride
self.backbone = backbone
in_channels = get_inchannels(self.backbone)
self.head = DANetHead(in_channels[0], num_classes, BatchNorm)
self.output_stride = output_stride
if freeze_bn:
self.freeze_bn()
def __init__(self, backbone,BatchNorm, output_stride, num_classes,freeze_bn=False):
super(DeepLab, self).__init__()
self.backbone = backbone
self.aspp = build_aspp(backbone, output_stride, BatchNorm)
self.decoder = build_decoder(num_classes, backbone, BatchNorm)
self.output_stride = output_stride
if freeze_bn:
self.freeze_bn()
def __init__(self,
backbone='resnet',
output_stride=16,
num_classes=21,
sync_bn=True,
freeze_bn=False):
super(DeepLab, self).__init__()
if 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)
if freeze_bn:
self.freeze_bn()
def __init__(self,
backbone,
BatchNorm,
output_stride,
num_classes,
freeze_bn=False):
super(DeepDran, self).__init__()
self.backbone = backbone
self.aspp = build_aspp(backbone, output_stride, BatchNorm)
self.decoder = build_decoder(num_classes, backbone, BatchNorm)
if (backbone in ["resnet50", "resnet101"]):
in_channels = 2048
in_channels_seg = 256
else:
raise NotImplementedError
self.head = DranHead(in_channels, num_classes, BatchNorm)
self.cls_seg = nn.Sequential(
nn.Dropout2d(0.1, False), nn.Conv2d(in_channels_seg, num_classes,
1))
self.output_stride = output_stride
if freeze_bn:
self.freeze_bn()