def __init__(self, batchNorm=False, other={}):
super(Regressor, self).__init__()
self.other = other
self.deconv1 = model_utils.conv(batchNorm, 128, 128, k=3, stride=1, pad=1)
self.deconv2 = model_utils.conv(batchNorm, 128, 128, k=3, stride=1, pad=1)
self.deconv3 = model_utils.deconv(128, 64)
self.est_normal= self._make_output(64, 3, k=3, stride=1, pad=1)
self.other = other
def __init__(self, conv_dim=64, init_zero_weights=False):
super(CycleGenerator, self).__init__()
# Define the encoder part of the generator (that extracts features from the input image)
self.conv1 = conv(3, conv_dim, 5, init_zero_weights=init_zero_weights)
self.conv2 = conv(conv_dim, conv_dim * 2, 5)
# Define the transformation part of the generator
self.resnet_block = ResnetBlock(conv_dim * 2)
# Define the decoder part of the generator (that builds up the output image from features)
self.upconv1 = upconv(conv_dim * 2, conv_dim, 5)
self.upconv2 = upconv(conv_dim, 3, 5, batch_norm=False)
def __init__(self, batchNorm, c_in, c_out=256):
super(FeatExtractor, self).__init__()
self.conv1 = model_utils.conv(batchNorm, c_in, 64, k=3, stride=2, pad=1)
self.conv2 = model_utils.conv(batchNorm, 64, 128, k=3, stride=2, pad=1)
self.conv3 = model_utils.conv(batchNorm, 128, 128, k=3, stride=1, pad=1)
self.conv4 = model_utils.conv(batchNorm, 128, 128, k=3, stride=2, pad=1)
self.conv5 = model_utils.conv(batchNorm, 128, 128, k=3, stride=1, pad=1)
self.conv6 = model_utils.conv(batchNorm, 128, 256, k=3, stride=2, pad=1)
self.conv7 = model_utils.conv(batchNorm, 256, 256, k=3, stride=1, pad=1)
def __init__(self, batchNorm=False, c_in=3, other={}):
super(FeatExtractor, self).__init__()
self.other = other
self.conv1 = model_utils.conv(batchNorm,
c_in,
64,
k=3,
stride=1,
pad=1)
self.conv2 = model_utils.conv(batchNorm, 64, 128, k=3, stride=2, pad=1)
self.conv3 = model_utils.conv(batchNorm,
128,
128,
k=3,
stride=1,
pad=1)
self.conv4 = model_utils.conv(batchNorm,
128,
256,
k=3,
stride=2,
pad=1)
self.conv5 = model_utils.conv(batchNorm,
256,
256,
k=3,
stride=1,
pad=1)
self.conv6 = model_utils.deconv(256, 128)
self.conv7 = model_utils.conv(batchNorm,
128,
128,
k=3,
stride=1,
pad=1)
def __init__(self, batchNorm, c_in, other):
super(Classifier, self).__init__()
self.conv1 = model_utils.conv(batchNorm, 512, 256, k=3, stride=1, pad=1)
self.conv2 = model_utils.conv(batchNorm, 256, 256, k=3, stride=2, pad=1)
self.conv3 = model_utils.conv(batchNorm, 256, 256, k=3, stride=2, pad=1)
self.conv4 = model_utils.conv(batchNorm, 256, 256, k=3, stride=2, pad=1)
self.other = other
self.dir_x_est = nn.Sequential(
model_utils.conv(batchNorm, 256, 64, k=1, stride=1, pad=0),
model_utils.outputConv(64, other['dirs_cls'], k=1, stride=1, pad=0))
self.dir_y_est = nn.Sequential(
model_utils.conv(batchNorm, 256, 64, k=1, stride=1, pad=0),
model_utils.outputConv(64, other['dirs_cls'], k=1, stride=1, pad=0))
self.int_est = nn.Sequential(
model_utils.conv(batchNorm, 256, 64, k=1, stride=1, pad=0),
model_utils.outputConv(64, other['ints_cls'], k=1, stride=1, pad=0))