class Executor:
def __init__(self, cf):
unet = Modified3DUNet(1, 1)
resnet = resnet34()
# device = torch.device('cpu')
unet.load_state_dict(torch.load(cf.pathToSegmentator))
resnet.load_state_dict(torch.load(cf.pathToClassifier))
unet.eval()
resnet.eval()
self.segmentator = Estimator(
unet,
save_folder='./experiments/unet_full_pipe_eval/',
cuda_device=0,
optimizer=Adam,
loss_fn=dice_loss)
self.classify = Estimator(
resnet,
save_folder='./experiments/res_full_pipe_eval/',
cuda_device=1,
optimizer=Adam,
loss_fn=torch.nn.CrossEntropyLoss())
self.pipe = Pipe(cf, self.classify, self.segmentator)
try:
shutil.rmtree(cf.save_path)
except Exception as e:
print(str(e))
try:
os.makedirs(cf.save_path)
print("Directory created")
except Exception as e:
print(str(e))
def unpack(self, pathToArchive, pathToConverted, numWorkers=3):
prep = Preprocess(pathToArchive, pathToConverted, numWorkers)
prep.start()
self.pipe.add_dataset(pathToConverted)
def start(self):
return self.pipe.start_inference()
