def __init__(self):
super(DeformNet, self).__init__()
self.conv1 = nn.Conv2d(1, 32, kernel_size=3, padding=1)
self.bn1 = nn.BatchNorm2d(32)
self.conv2 = nn.Conv2d(32, 64, kernel_size=3, padding=1)
self.bn2 = nn.BatchNorm2d(64)
self.conv3 = nn.Conv2d(64, 128, kernel_size=3, padding=1)
self.bn3 = nn.BatchNorm2d(128)
self.offsets = nn.Conv2d(128, 18, kernel_size=3, padding=1)
self.conv4 = DeformConv2D(128, 128, kernel_size=3, padding=1)
self.bn4 = nn.BatchNorm2d(128)
self.classifier = nn.Linear(128, 10)
def __init__(self):
super(DeformNet, self).__init__()
self.conv1 = nn.Conv2d(3, 32, kernel_size=3, padding=1)
self.bn1 = nn.BatchNorm2d(32)
self.conv2 = nn.Conv2d(32, 64, kernel_size=3, padding=1)
self.bn2 = nn.BatchNorm2d(64)
self.conv3 = nn.Conv2d(64, 128, kernel_size=3, padding=1)
self.bn3 = nn.BatchNorm2d(128)
# the 18 is the 2 * 3 * 3 (3 * 3 is the size of the kernel) - multiply with 2 because it generate
# the offsets coordinates for each of the 3 * 3 kernel
self.offsets = nn.Conv2d(128, 18, kernel_size=3, padding=1) # first step in DCN process
self.conv4 = DeformConv2D(128, 128, kernel_size=3, padding=1)
self.bn4 = nn.BatchNorm2d(128)
self.classifier = nn.Linear(128, 10)
def __init__(self):
super(Fine, self).__init__()
self.conv1_1 = nn.Conv2d(1, 64, 3, padding=1)
self.relu1_1 = nn.ReLU(inplace=True)
self.conv1_2 = nn.Conv2d(64, 64, 3, padding=1)
self.relu1_2 = nn.ReLU(inplace=True)
self.pool1 = nn.MaxPool2d(2, stride=2, ceil_mode=True) # 1/2
# conv2
self.conv2_1 = nn.Conv2d(64, 128, 3, padding=1)
self.relu2_1 = nn.ReLU(inplace=True)
self.conv2_2 = nn.Conv2d(128, 128, 3, padding=1)
self.relu2_2 = nn.ReLU(inplace=True)
self.pool2 = nn.MaxPool2d(2, stride=2, ceil_mode=True) # 1/4
# conv3
self.conv3_1 = nn.Conv2d(128, 256, 3, padding=1)
self.relu3_1 = nn.ReLU(inplace=True)
self.conv3_2 = nn.Conv2d(256, 256, 3, padding=1)
self.relu3_2 = nn.ReLU(inplace=True)
self.conv3_3 = nn.Conv2d(256, 256, 3, padding=1)
self.relu3_3 = nn.ReLU(inplace=True)
self.pool3 = nn.MaxPool2d(2, stride=2, ceil_mode=True) # 1/8
# conv4
self.conv4_1 = nn.Conv2d(256, 512, 3, padding=1)
self.relu4_1 = nn.ReLU(inplace=True)
self.conv4_2 = nn.Conv2d(512, 512, 3, padding=1)
self.relu4_2 = nn.ReLU(inplace=True)
self.conv4_3 = nn.Conv2d(512, 512, 3, padding=1)
self.relu4_3 = nn.ReLU(inplace=True)
self.pool4 = nn.MaxPool2d(2, stride=2, ceil_mode=True) # 1/16
#conv5
self.conv5_1 = nn.Conv2d(512, 512, 3, padding=1)
self.relu5_1 = nn.ReLU(inplace=True)
self.conv5_2 = nn.Conv2d(512, 512, 3, padding=1)
self.relu5_2 = nn.ReLU(inplace=True)
self.conv5_3 = nn.Conv2d(512, 512, 3, padding=1)
self.relu5_3 = nn.ReLU(inplace=True)
self.conv6_1 = nn.Conv2d(512, 256, 3, padding=1)
self.relu6_1 = nn.ReLU(inplace=True)
self.conv6_2 = nn.Conv2d(256, 128, 3, padding=1)
self.relu6_2 = nn.ReLU(inplace=True)
self.conv6_3 = nn.Conv2d(128, 64, 3, padding=1)
self.relu6_3 = nn.ReLU(inplace=True)
self.conv7_1 = nn.Conv2d(64, 32, 3, padding=1)
self.relu7_1 = nn.ReLU(inplace=True)
self.conv7_2 = nn.Conv2d(32, 16, 3, padding=1)
self.relu7_2 = nn.ReLU(inplace=True)
self.conv7_3 = nn.Conv2d(16, 8, 3, padding=1)
self.relu7_3 = nn.ReLU(inplace=True)
self.conv8_1 = nn.Conv2d(8, 4, 3, padding=1)
self.relu8_1 = nn.ReLU(inplace=True)
self.conv8_2 = nn.Conv2d(4, 2, 3, padding=1)
self.relu8_2 = nn.ReLU(inplace=True)
self.conv8_3 = nn.Conv2d(2, 1, 3, padding=1)
self.relu8_3 = nn.ReLU(inplace=True)
self.offsets4 = nn.Conv2d(512, 18, kernel_size=3, padding=1)
self.conv4 = DeformConv2D(512, 1, 3, padding=1)
self.relu4 = nn.ReLU(inplace=True)
self.offsets5 = nn.Conv2d(512, 18, kernel_size=3, padding=1)
self.conv5 = DeformConv2D(512, 1, 3, padding=1)
self.relu5 = nn.ReLU(inplace=True)
self.offsets6 = nn.Conv2d(64, 18, kernel_size=3, padding=1)
self.conv6 = DeformConv2D(64, 1, 3, padding=1)
self.relu6 = nn.ReLU(inplace=True)
self.offsets7 = nn.Conv2d(8, 18, kernel_size=3, padding=1)
self.conv7 = DeformConv2D(8, 1, 3, padding=1)
self.relu7 = nn.ReLU(inplace=True)
self.w4 = nn.Parameter(torch.ones(1))
self.w5 = nn.Parameter(torch.ones(1))
self.w6 = nn.Parameter(torch.ones(1))
self.w7 = nn.Parameter(torch.ones(1))
self.w8 = nn.Parameter(torch.ones(1))
def __init__(self, load_weights=False):
super(ASPDNet, self).__init__()
self.seen = 0
## frontend feature extraction
self.frontend_feat = [
64, 64, 'M', 128, 128, 'M', 256, 256, 256, 'M', 512, 512, 512
]
self.mid_feat = [512, 512, 512]
self.frontend = make_layers(self.frontend_feat)
'''
# CBAM module (convolution block attention module)
cite as "CBAM: Convolutional Block Attention Module, 2018 ECCV"
'''
self.planes = self.frontend_feat[-1]
self.ca = ChannelAttention(self.planes)
self.sa = SpatialAttention()
'''
dilation convolution (Spatial Pyramid Module)
cite as "Scale Pyramid Network for Crowd Counting, 2019 WACV"
'''
self.conv4_3_1 = nn.Conv2d(512,
512,
kernel_size=3,
padding=2,
dilation=2)
self.conv4_3_2 = nn.Conv2d(512,
512,
kernel_size=3,
padding=4,
dilation=4)
self.conv4_3_3 = nn.Conv2d(512,
512,
kernel_size=3,
padding=8,
dilation=8)
self.conv4_3_4 = nn.Conv2d(512,
512,
kernel_size=3,
padding=12,
dilation=12)
# self.conv4 = [self.conv4_3_1, self.conv4_3_2, self.conv4_3_3, self.conv4_3_4]
self.conv5 = nn.Conv2d(2048, 512, kernel_size=1)
'''
convolution layers
'''
self.mid_end = make_layers(
self.mid_feat,
in_channels=512,
)
'''
deformable convolution network
cite as "Deformable Convolutional Networks, 2017 ICCV"
'''
self.offset1 = nn.Conv2d(512, 18, kernel_size=3, padding=1)
self.conv6_1 = DeformConv2D(512, 256, kernel_size=3, padding=1)
self.bn6_1 = nn.BatchNorm2d(256)
self.offset2 = nn.Conv2d(256, 18, kernel_size=3, padding=1)
self.conv6_2 = DeformConv2D(256, 128, kernel_size=3, padding=1)
self.bn6_2 = nn.BatchNorm2d(128)
self.offset3 = nn.Conv2d(128, 18, kernel_size=3, padding=1)
self.conv6_3 = DeformConv2D(128, 64, kernel_size=3, padding=1)
self.bn6_3 = nn.BatchNorm2d(64)
self.output_layer = nn.Conv2d(64, 1, kernel_size=1)
if not load_weights:
mod = models.vgg16(pretrained=True)
self._initialize_weights()
for i in range(len(self.frontend.state_dict().items())):
list(self.frontend.state_dict().items())[i][1].data[:] = list(
mod.state_dict().items())[i][1].data[:]
def __init__(self, in_c, out_c): #inc, outc 이거 알아서 파라미터로 받아보기
super(deformable_filter, self).__init__()
self.offset = nn.Conv2d(in_c, 18, kernel_size=3, padding=1)
self.deform_conv = DeformConv2D(in_c, out_c, kernel_size=3, padding=1)
init_conv_offset(self.offset)