unet = Unet(1, 1)
sample = Sample()
if torch.cuda.is_available():
unet = unet.cuda()
criterion = nn.BCELoss()
# optimizer = optim.RMSprop(unet.parameters(),lr=0.001)
lr = random.random() * 0.0005 + 0.0000005
optimizer = optim.Adam(unet.parameters(), lr=lr, betas=(0.9, 0.999))
cout1 = 0
cout2 = 0
cout3 = 0
dice = 0
for epoch in range(50000):
while True:
try:
imgs, gts = sample.get_batch(4)
break
except:
pass
imgs = np.array(imgs)
gts = np.array(gts)
gts = np.expand_dims(gts, axis=1)
imgs = torch.Tensor(imgs)
gts = torch.Tensor(gts)
if torch.cuda.is_available():
imgs = imgs.cuda()
gts = gts.cuda()
else:
imgs = Variable(imgs)
gts = Variable(gts)
out = unet(imgs)
