def __init__(self, pretrained=False, freeze=False): super(CRAFT, self).__init__() #basenet self.basenet = vgg16_bn(pretrained, freeze) #U-net self.upconv1 = double_conv(1024, 512, 256) self.upconv2 = double_conv(512, 256, 128) self.upconv3 = double_conv(256, 128, 64) self.upconv4 = double_conv(128, 64, 32) num_class = 2 self.conv_cls = nn.Sequential( nn.Conv2d(32, 32, kernel_size=3, padding=1), nn.ReLU(inplace=True), nn.Conv2d(32, 32, kernel_size=3, padding=1), nn.ReLU(inplace=True), nn.Conv2d(32, 16, kernel_size=3, padding=1), nn.ReLU(inplace=True), nn.Conv2d(16, 16, kernel_size=1), nn.ReLU(inplace=True), nn.Conv2d(16, num_class, kernel_size=1), ) init_weights(self.upconv1.modules()) init_weights(self.upconv2.modules()) init_weights(self.upconv3.modules()) init_weights(self.upconv4.modules()) init_weights(self.conv_cls.modules())
def __init__(self, pretrained=False, freeze=False): super(CRAFT, self).__init__() """ Base network """ self.basenet = vgg16_bn(pretrained, freeze) """ U network """ self.upconv1 = double_conv(1024, 512, 256) self.upconv2 = double_conv(512, 256, 128) self.upconv3 = double_conv(256, 128, 64) self.upconv4 = double_conv(128, 64, 32) self.roialign_s16 = ROIAlign((1, 1), 1 / 16, 0) self.roialign_s8 = ROIAlign((1, 1), 1 / 8, 0) self.roialign_s4 = ROIAlign((1, 1), 1 / 4, 0) self.roialign_s2 = ROIAlign((1, 1), 1 / 2, 0) num_class = 2 self.conv_cls = nn.Sequential( nn.Conv2d(32, 32, kernel_size=3, padding=1), nn.ReLU(inplace=True), nn.Conv2d(32, 32, kernel_size=3, padding=1), nn.ReLU(inplace=True), nn.Conv2d(32, 16, kernel_size=3, padding=1), nn.ReLU(inplace=True), nn.Conv2d(16, 16, kernel_size=1), nn.ReLU(inplace=True), nn.Conv2d(16, num_class, kernel_size=1), ) init_weights(self.upconv1.modules()) init_weights(self.upconv2.modules()) init_weights(self.upconv3.modules()) init_weights(self.upconv4.modules()) init_weights(self.conv_cls.modules())
def __init__(self): super(RefineNet, self).__init__() self.last_conv = nn.Sequential( nn.Conv2d(34, 64, kernel_size=3, padding=1), nn.BatchNorm2d(64), nn.ReLU(inplace=True), nn.Conv2d(64, 64, kernel_size=3, padding=1), nn.BatchNorm2d(64), nn.ReLU(inplace=True), nn.Conv2d(64, 64, kernel_size=3, padding=1), nn.BatchNorm2d(64), nn.ReLU(inplace=True)) self.aspp1 = nn.Sequential( nn.Conv2d(64, 128, kernel_size=3, dilation=6, padding=6), nn.BatchNorm2d(128), nn.ReLU(inplace=True), nn.Conv2d(128, 128, kernel_size=1), nn.BatchNorm2d(128), nn.ReLU(inplace=True), nn.Conv2d(128, 1, kernel_size=1)) self.aspp2 = nn.Sequential( nn.Conv2d(64, 128, kernel_size=3, dilation=12, padding=12), nn.BatchNorm2d(128), nn.ReLU(inplace=True), nn.Conv2d(128, 128, kernel_size=1), nn.BatchNorm2d(128), nn.ReLU(inplace=True), nn.Conv2d(128, 1, kernel_size=1)) self.aspp3 = nn.Sequential( nn.Conv2d(64, 128, kernel_size=3, dilation=18, padding=18), nn.BatchNorm2d(128), nn.ReLU(inplace=True), nn.Conv2d(128, 128, kernel_size=1), nn.BatchNorm2d(128), nn.ReLU(inplace=True), nn.Conv2d(128, 1, kernel_size=1)) self.aspp4 = nn.Sequential( nn.Conv2d(64, 128, kernel_size=3, dilation=24, padding=24), nn.BatchNorm2d(128), nn.ReLU(inplace=True), nn.Conv2d(128, 128, kernel_size=1), nn.BatchNorm2d(128), nn.ReLU(inplace=True), nn.Conv2d(128, 1, kernel_size=1)) init_weights(self.last_conv.modules()) init_weights(self.aspp1.modules()) init_weights(self.aspp2.modules()) init_weights(self.aspp3.modules()) init_weights(self.aspp4.modules())
def __init__(self, pretrained=False, freeze=False): super(CRAFT, self).__init__() """ Base network """ self.basenet = vgg16_bn(pretrained, freeze) """ U network """ self.upconv1 = double_conv(1024, 512, 256) self.upconv2 = double_conv(512, 256, 128) self.upconv3 = double_conv(256, 128, 64) self.upconv4 = double_conv(128, 64, 32) #num_class = 2 self.conv_cls = Capsulenet.CapsNet() init_weights(self.upconv1.modules()) init_weights(self.upconv2.modules()) init_weights(self.upconv3.modules()) init_weights(self.upconv4.modules())