def __init__(self):
# ID and Name
self.id = 205
self.experiment_name = "pancreas_2D_unet_{}".format(self.id)
self.debug = False
# System
self.checkpointsBasePath = "./checkpoints/"
self.checkpointsBasePathMod = self.checkpointsBasePath + self.experiment_name + '/'
self.datapath = "/local/DEEPLEARNING/TCIA/"
self.split = 0
# GPU
self.gpu = '1'
os.environ["CUDA_VISIBLE_DEVICES"] = self.gpu
# Model
self.channels = [64, 128, 256, 512, 1024]
self.channels = [int(x) for x in self.channels]
self.net = UNet(filters=self.channels,
n_classes=2,
in_channels=1,
dim='2d',
bn=False,
up_mode='deconv')
self.n_parameters = count_parameters(self.net)
print("N PARAMS : {}".format(self.n_parameters))
self.model_path = './checkpoints/models/pancreas2D_unet_kaiming_nobn_deconv.pth'
self.load_model()
self.n_classes = 2
# self.transform = tf.Compose([
# tf.RandomAffine(degrees = 10,
# scale = (0.9,1.1),
# translate = (0.1, 0.1)),
# tf.ToTensor(),
# ])
self.transform = True
# Training
self.start_epoch = 0
self.epoch = 20
self.loss = torch.nn.CrossEntropyLoss()
self.batchsize = 3
self.lr_rate = 1e-4
self.final_lr_rate = 1e-5
self.optimizer = optim.Adam(self.net.parameters(), lr=self.lr_rate)
self.optimizer.zero_grad()
self.validate_every_k_epochs = 1
# self.lr_scheduler = get_scheduler(self.optimizer, "constant", self.lr_rate)
# self.lr_scheduler = optim.lr_scheduler.ExponentialLR(self.optimizer, 0.99)
self.decay = (self.lr_rate / self.final_lr_rate - 1) / self.epoch
self.lr_scheduler = get_scheduler(self.optimizer, "po", {
'lr': self.lr_rate,
'decay': self.decay
})
# Other
self.classes_name = ['background', 'pancreas']
def __init__(self):
# ID and Name
self.id = 21
self.experiment_name = "pancreas_2D_unet_{}".format(self.id)
self.debug = False
# System
self.checkpointsBasePath = "./checkpoints/"
self.checkpointsBasePathMod = self.checkpointsBasePath + self.experiment_name + '/'
self.datapath = "/local/DEEPLEARNING/TCIA/"
self.split = 0
# GPU
self.gpu = '0'
os.environ["CUDA_VISIBLE_DEVICES"] = self.gpu
# Model
self.channels = [64, 128, 256, 512, 1024]
self.channels = [int(x) for x in self.channels]
self.net = UNet(filters=self.channels,
n_classes=2,
in_channels=1,
dim='2d')
self.n_parameters = count_parameters(self.net)
print("N PARAMS : {}".format(self.n_parameters))
self.model_path = './checkpoints/models/pancreas2D_unet.pth'
self.load_model()
self.n_classes = 2
self.transform = tf.Compose([
tf.RandomAffine(degrees=5,
scale=(0.95, 1.05),
translate=(0.05, 0.05)),
tf.ToTensor(),
])
# Training
self.start_epoch = 0
self.epoch = 300
self.loss = torch.nn.CrossEntropyLoss()
self.batchsize = 8
self.lr_rate = 1e-4
self.optimizer = optim.Adam(self.net.parameters(),
lr=self.lr_rate,
weight_decay=0)
self.optimizer.zero_grad()
self.validate_every_k_epochs = 1
# self.lr_scheduler = get_scheduler(self.optimizer, "constant", self.lr_rate)
self.lr_scheduler = optim.lr_scheduler.ExponentialLR(
self.optimizer, 0.96)
# Other
self.classes_name = ['background', 'pancreas']
def __init__(self, patch_size, filters, n_classes=2, in_channels=1):
super(Patched3DUNet, self).__init__()
self.in_channels = in_channels
self.dim = '3d'
self.filters = filters
# self.patch_size = patch_size
self.ps_h, self.ps_w, self.ps_d = patch_size
self.n_classes = n_classes
self.unet = UNet(filters=self.filters,
n_classes=self.n_classes,
in_channels=self.in_channels,
dim=self.dim)