def __init__(self):
super(Generator, self).__init__()
self.conv1 = Conv2d(8, 32, kernel_size=7, stride=1)
self.conv2 = Conv2d(32, 64, kernel_size=3, stride=2)
self.conv3 = Conv2d(64, 128, kernel_size=3, stride=2)
self.conv4 = Conv2d(32, 3, kernel_size=7, stride=1)
self.bn1 = nn.BatchNorm2d(32, eps=0.001, track_running_stats=True)
self.bn2 = nn.BatchNorm2d(64, eps=0.001, track_running_stats=True)
self.bn3 = nn.BatchNorm2d(128, eps=0.001, track_running_stats=True)
self.bn4 = nn.BatchNorm2d(64, eps=0.001, track_running_stats=True)
self.bn5 = nn.BatchNorm2d(32, eps=0.001, track_running_stats=True)
self.repeat_blocks = self._make_repeat_blocks(BasicBlock(128, 128), 6)
self.deconv1 = nn.ConvTranspose2d(128,
64,
kernel_size=3,
stride=2,
padding=1)
self.deconv2 = nn.ConvTranspose2d(64,
32,
kernel_size=3,
stride=2,
padding=0,
output_padding=1)
self.relu = nn.ReLU()
self.tanh = nn.Tanh()
def __init__(self):
super(PatchDiscriminator, self).__init__()
self.lrelu = nn.LeakyReLU(0.2)
self.conv1 = Conv2d(3, 64, kernel_size=4, stride=2)
self.conv2 = Conv2d(69, 128, kernel_size=4, stride=2)
self.bn2 = nn.BatchNorm2d(128, eps=0.001, track_running_stats=True)
self.conv3 = Conv2d(128, 256, kernel_size=4, stride=2)
self.bn3 = nn.BatchNorm2d(256, eps=0.001, track_running_stats=True)
self.conv4 = Conv2d(256, 512, kernel_size=4, stride=2)
self.bn4 = nn.BatchNorm2d(512, eps=0.001, track_running_stats=True)
self.conv5 = Conv2d(512, 512, kernel_size=4, stride=2)
def __init__(self, pretrainded=False, modelpath=None):
super(PatchDiscriminator, self).__init__()
self.lrelu = nn.LeakyReLU(0.2)
self.conv1 = Conv2d(3, 64, kernel_size=4, stride=2)
self.conv2 = Conv2d(69, 128, kernel_size=4, stride=2)
self.bn2 = nn.BatchNorm2d(128, eps=0.001, track_running_stats=True)
self.conv3 = Conv2d(128, 256, kernel_size=4, stride=2)
self.bn3 = nn.BatchNorm2d(256, eps=0.001, track_running_stats=True)
self.conv4 = Conv2d(256, 512, kernel_size=4, stride=2)
self.bn4 = nn.BatchNorm2d(512, eps=0.001, track_running_stats=True)
self.conv5 = Conv2d(512, 512, kernel_size=4, stride=2)
if pretrainded:
self.load_pretrained_params(modelpath)
def __init__(self,pretrainded=False,modelpath=None):
super(AgeAlexNet, self).__init__()
assert pretrainded is False or modelpath is not None,"pretrain model need to be specified"
self.features = nn.Sequential(
Conv2d(3, 96, kernel_size=11, stride=4),
nn.ReLU(inplace=True),
nn.MaxPool2d(kernel_size=3, stride=2),
nn.LocalResponseNorm(2,2e-5,0.75),
Conv2d(96, 256, kernel_size=5, stride=1,groups=2),
nn.ReLU(inplace=True),
nn.MaxPool2d(kernel_size=3, stride=2),
nn.LocalResponseNorm(2, 2e-5, 0.75),
Conv2d(256, 384, kernel_size=3, stride=1),
nn.ReLU(inplace=True),
Conv2d(384, 384, kernel_size=3,stride=1,groups=2),
nn.ReLU(inplace=True),
Conv2d(384, 256, kernel_size=3,stride=1,groups=2),
nn.ReLU(inplace=True),
nn.MaxPool2d(kernel_size=3, stride=2),
)
self.age_classifier=nn.Sequential(
nn.Dropout(),
nn.Linear(256 * 6 * 6, 4096),
nn.ReLU(inplace=True),
nn.Dropout(),
nn.Linear(4096, 4096),
nn.ReLU(inplace=True),
nn.Linear(4096, 5),
)
if pretrainded is True:
self.load_pretrained_params(modelpath)
self.Conv3_feature_module=nn.Sequential()
self.Conv4_feature_module=nn.Sequential()
self.Conv5_feature_module=nn.Sequential()
self.Pool5_feature_module=nn.Sequential()
for x in range(10):
self.Conv3_feature_module.add_module(str(x), self.features[x])
for x in range(10,12):
self.Conv4_feature_module.add_module(str(x),self.features[x])
for x in range(12,14):
self.Conv5_feature_module.add_module(str(x),self.features[x])
for x in range(14,15):
self.Pool5_feature_module.add_module(str(x),self.features[x])
def __init__(self):
super(RPN, self).__init__()
self.features = VGG16(bn=False)
self.conv1 = Conv2d(512, 512, 3, same_padding=True)
self.score_conv = Conv2d(512,
len(self.anchor_scales) * 3 * 2,
1,
relu=False,
same_padding=False)
self.bbox_conv = Conv2d(512,
len(self.anchor_scales) * 3 * 4,
1,
relu=False,
same_padding=False)
# loss
self.cross_entropy = None
self.los_box = None