def __init__(self, num_classes, pretrained=True, phase='train'):
super(FCN8VGG, self).__init__()
vgg = models.vgg19_bn()
if pretrained:
vgg.load_state_dict(torch.load(vgg19_bn_path))
features = list(vgg.features.children())
self.features3 = nn.Sequential(*features[0:27])
self.features4_4 = nn.Sequential(*features[27:39])
self.features4 = nn.Sequential(*features[39:40])
self.features5 = nn.Sequential(*features[40:])
self.features5_test = nn.Sequential(*features[40:52])
self.fconv3 = nn.Conv2d(256, num_classes, kernel_size=1)
self.fconv4 = nn.Conv2d(512, num_classes, kernel_size=1)
self.fconv5 = nn.Sequential(
nn.Conv2d(512, 2048, kernel_size=7),
nn.ReLU(inplace=True),
nn.Dropout(),
nn.Conv2d(2048, 2048, kernel_size=1),
nn.ReLU(inplace=True),
nn.Dropout(),
nn.Conv2d(2048, num_classes, kernel_size=1)
)
initialize_weights(self.fconv3, self.fconv4, self.fconv5)
self.ssd_conv5_1 = nn.Conv2d(512, 512, kernel_size=3, padding=1, dilation=1)
def __init__(self, num_classes, pretrained=True):
super(FCN8ResNet, self).__init__()
res = models.resnet152()
if pretrained:
res.load_state_dict(torch.load(res152_path))
self.features3 = nn.Sequential(res.conv1, res.bn1, res.relu,
res.maxpool, res.layer1, res.layer2)
# pdb.set_trace()
self.features4 = res.layer3
self.features5 = res.layer4
if cfg.TRAIN.LOC_P:
loc_tmp = np.arange(0, 2, 2.0 / 64) - 1.0
loc_x = np.tile(loc_tmp, (64, 1))
loc_y = loc_x.transpose()
loc = np.concatenate((loc_x[None, :, :], loc_x[None, :, :]),
axis=0)[None, :, :, :]
loc_input = np.repeat(loc, cfg.TRAIN.IMG_BATCH_SIZE, axis=0)
loc_input = torch.from_numpy(loc_input.astype(np.float32))
self.loc_input = Variable(loc_input.cuda())
self.fconv3 = nn.Conv2d(514, num_classes, kernel_size=1)
else:
self.fconv3 = nn.Conv2d(512, num_classes, kernel_size=1)
self.fconv4 = nn.Conv2d(2048, num_classes, kernel_size=1)
self.fconv5 = nn.Conv2d(2560, num_classes, kernel_size=7)
initialize_weights(self.fconv3, self.fconv4, self.fconv5)
self.ssd_conv5 = nn.Conv2d(1024,
512,
kernel_size=3,
padding=1,
dilation=1)
def __init__(self, num_classes, pretrained=True):
super(FCN8ResNet, self).__init__()
res = models.resnet152()
if pretrained:
res.load_state_dict(torch.load(res152_path))
self.features3 = nn.Sequential(
res.conv1, res.bn1, res.relu, res.maxpool, res.layer1, res.layer2
)
# pdb.set_trace()
self.features4 = res.layer3
self.features5 = res.layer4
self.fconv3 = nn.Conv2d(512, num_classes, kernel_size=1)
self.fconv4 = nn.Conv2d(2048, num_classes, kernel_size=1)
self.fconv5 = nn.Conv2d(2560, num_classes, kernel_size=7)
initialize_weights(self.fconv3, self.fconv4, self.fconv5)
self.ssd_conv5 = nn.Conv2d(1024, 512, kernel_size=3, padding=1, dilation=1)
def __init__(self, num_classes, pretrained=True, phase='train'):
super(FCN8VGG, self).__init__()
vgg = models.vgg19_bn()
if pretrained:
vgg.load_state_dict(torch.load(vgg19_bn_path))
features = list(vgg.features.children())
self.features3 = nn.Sequential(*features[0:27])
self.features4_4 = nn.Sequential(*features[27:39])
self.features4 = nn.Sequential(*features[39:40])
self.features5 = nn.Sequential(*features[40:])
self.features5_test = nn.Sequential(*features[40:52])
if cfg.TRAIN.LOC_P:
loc_tmp = np.arange(0, 2, 2.0 / 64) - 1.0
loc_x = np.tile(loc_tmp, (64, 1))
loc_y = loc_x.transpose()
loc = np.concatenate((loc_x[None, :, :], loc_x[None, :, :]),
axis=0)[None, :, :, :]
loc_input = np.repeat(loc, cfg.TRAIN.IMG_BATCH_SIZE, axis=0)
loc_input = torch.from_numpy(loc_input.astype(np.float32))
self.loc_input = Variable(loc_input.cuda())
self.fconv3 = nn.Conv2d(258, num_classes, kernel_size=1)
else:
self.fconv3 = nn.Conv2d(256, num_classes, kernel_size=1)
self.fconv4 = nn.Conv2d(1536, num_classes, kernel_size=1)
self.fconv5 = nn.Sequential(
nn.Conv2d(1024, 2048, kernel_size=7), nn.ReLU(inplace=True),
nn.Dropout(), nn.Conv2d(2048, 2048, kernel_size=1),
nn.ReLU(inplace=True), nn.Dropout(),
nn.Conv2d(2048, num_classes, kernel_size=1))
initialize_weights(self.fconv3, self.fconv4, self.fconv5)
self.ssd_conv5 = nn.Conv2d(512,
512,
kernel_size=3,
padding=1,
dilation=1)
def __init__(self, num_classes, pretrained=True, phase='train'):
super(FCN8VGG, self).__init__()
vgg = models.vgg19_bn()
if pretrained:
vgg.load_state_dict(torch.load(vgg19_bn_path))
features = list(vgg.features.children())
self.features3_4 = nn.Sequential(
*features[0:26]) # for FCN 256,128,128
# self.features3 = nn.Sequential(*features[26:27]) # for FCN 256,64,64
self.features4_4 = nn.Sequential(*features[26:39]) # for FCN 256,64,64
# self.features4 = nn.Sequential(*features[39:40]) # for FCN 512, 32, 32
self.features5_4 = nn.Sequential(
*features[39:52]) # for FCN 512, 32, 32
# self.features5 = nn.Sequential(*features[52:]) # for FCN 512, 16, 16
self.fconv3 = nn.Conv2d(256, num_classes, kernel_size=1)
self.fconv4 = nn.Conv2d(512, num_classes, kernel_size=1)
self.fconv5 = nn.Sequential(
nn.Conv2d(1536, 2048, kernel_size=7), nn.ReLU(inplace=True),
nn.Dropout(), nn.Conv2d(2048, 2048, kernel_size=1),
nn.ReLU(inplace=True), nn.Dropout(),
nn.Conv2d(2048, num_classes, kernel_size=1))
initialize_weights(self.fconv3, self.fconv4, self.fconv5)