Error when trying to use 2D nifti data

[covasquezv]

Hi Dominik, I was using MIScnn to train a U-Net with patches from 3D nifti data and I didn't had any issues related to data. But now I've change the data to 2D slices (also nifti) an during first epoch I had the following error related to loss functions:

Epoch 1/1000
Traceback (most recent call last):
  File "scripts/run_miscnn_2D.py", line 145, in <module>
    callbacks=[cb_lr, cb_es, cb_tb, cb_cl, cb_ckp])
  File "/home/ubuntu/.local/lib/python3.6/site-packages/miscnn/neural_network/model.py", line 201, in evaluate
    max_queue_size=self.batch_queue_size)
  File "/home/ubuntu/.local/lib/python3.6/site-packages/tensorflow/python/keras/engine/training.py", line 66, in _method_wrapper
    return method(self, *args, **kwargs)
  File "/home/ubuntu/.local/lib/python3.6/site-packages/tensorflow/python/keras/engine/training.py", line 848, in fit
    tmp_logs = train_function(iterator)
  File "/home/ubuntu/.local/lib/python3.6/site-packages/tensorflow/python/eager/def_function.py", line 580, in __call__
    result = self._call(*args, **kwds)
  File "/home/ubuntu/.local/lib/python3.6/site-packages/tensorflow/python/eager/def_function.py", line 627, in _call
    self._initialize(args, kwds, add_initializers_to=initializers)
  File "/home/ubuntu/.local/lib/python3.6/site-packages/tensorflow/python/eager/def_function.py", line 506, in _initialize
    *args, **kwds))
  File "/home/ubuntu/.local/lib/python3.6/site-packages/tensorflow/python/eager/function.py", line 2446, in _get_concrete_function_internal_garbage_collected
    graph_function, _, _ = self._maybe_define_function(args, kwargs)
  File "/home/ubuntu/.local/lib/python3.6/site-packages/tensorflow/python/eager/function.py", line 2777, in _maybe_define_function
    graph_function = self._create_graph_function(args, kwargs)
  File "/home/ubuntu/.local/lib/python3.6/site-packages/tensorflow/python/eager/function.py", line 2667, in _create_graph_function
    capture_by_value=self._capture_by_value),
  File "/home/ubuntu/.local/lib/python3.6/site-packages/tensorflow/python/framework/func_graph.py", line 981, in func_graph_from_py_func
    func_outputs = python_func(*func_args, **func_kwargs)
  File "/home/ubuntu/.local/lib/python3.6/site-packages/tensorflow/python/eager/def_function.py", line 441, in wrapped_fn
    return weak_wrapped_fn().__wrapped__(*args, **kwds)
  File "/home/ubuntu/.local/lib/python3.6/site-packages/tensorflow/python/framework/func_graph.py", line 968, in wrapper
    raise e.ag_error_metadata.to_exception(e)
ValueError: in user code:

    /home/ubuntu/.local/lib/python3.6/site-packages/tensorflow/python/keras/engine/training.py:571 train_function  *
        outputs = self.distribute_strategy.run(
    /home/ubuntu/.local/lib/python3.6/site-packages/tensorflow/python/distribute/distribute_lib.py:951 run  **
        return self._extended.call_for_each_replica(fn, args=args, kwargs=kwargs)
    /home/ubuntu/.local/lib/python3.6/site-packages/tensorflow/python/distribute/distribute_lib.py:2290 call_for_each_replica
        return self._call_for_each_replica(fn, args, kwargs)
    /home/ubuntu/.local/lib/python3.6/site-packages/tensorflow/python/distribute/distribute_lib.py:2649 _call_for_each_replica
        return fn(*args, **kwargs)
    /home/ubuntu/.local/lib/python3.6/site-packages/tensorflow/python/keras/engine/training.py:533 train_step  **
        y, y_pred, sample_weight, regularization_losses=self.losses)
    /home/ubuntu/.local/lib/python3.6/site-packages/tensorflow/python/keras/engine/compile_utils.py:205 __call__
        loss_value = loss_obj(y_t, y_p, sample_weight=sw)
    /home/ubuntu/.local/lib/python3.6/site-packages/tensorflow/python/keras/losses.py:143 __call__
        losses = self.call(y_true, y_pred)
    /home/ubuntu/.local/lib/python3.6/site-packages/tensorflow/python/keras/losses.py:246 call
        return self.fn(y_true, y_pred, **self._fn_kwargs)
    /home/ubuntu/.local/lib/python3.6/site-packages/miscnn/neural_network/metrics.py:126 tversky_crossentropy
        crossentropy = K.categorical_crossentropy(y_truth, y_pred)
    /home/ubuntu/.local/lib/python3.6/site-packages/tensorflow/python/keras/backend.py:4561 categorical_crossentropy
        target.shape.assert_is_compatible_with(output.shape)
    /home/ubuntu/.local/lib/python3.6/site-packages/tensorflow/python/framework/tensor_shape.py:1117 assert_is_compatible_with
        raise ValueError("Shapes %s and %s are incompatible" % (self, other))

    ValueError: Shapes (None, None, None, None) and (None, None, None, None, 2) are incompatible

I'm thinking is something related to my architecture configuration or the metrics definition, but I don't understand what. Here's my pipeline:

interface = NIFTI_interface(channels=1, classes=2)
# Create Data IO object to load and write samples in the file structure
data_io = Data_IO(interface, input_path="data_2D", delete_batchDir=False)

# Access all available samples in our file structure
sample_list = data_io.get_indiceslist()
sample_list.sort()

# Create and configure the Data Augmentation class
data_aug = Data_Augmentation(cycles=1, scaling=True, rotations=True,
                             elastic_deform=True, mirror=True,
                             brightness=True, contrast=True, gamma=True,
                             gaussian_noise=True)

# Create a clipping Subfunction to the lung window of CTs (-1250 and 250)
sf_clipping = Clipping(min=-1250, max=250)
# Create a pixel value normalization Subfunction for z-score scaling
sf_zscore = Normalization(mode="z-score")

sf_resize = Resize((512, 512))

# Assemble Subfunction classes into a list
sf = [sf_clipping, sf_zscore, sf_resize]

# Create and configure the Preprocessor class
pp = Preprocessor(data_io, data_aug=data_aug, batch_size=8, subfunctions=sf,
                  prepare_subfunctions=True, prepare_batches=False,
                  analysis="fullimage")

# Initialize the Architecture
unet_standard = Architecture(depth=3, activation="softmax",
                             batch_normalization=True)

# Create the Neural Network model
model = Neural_Network(preprocessor=pp, architecture=unet_standard,
                       loss=tversky_crossentropy,
                       metrics=[tversky_loss, dice_soft, dice_crossentropy],
                       batch_queue_size=3, workers=3, learninig_rate=0.001)

# Define Callbacks
cb_lr = ReduceLROnPlateau(monitor='val_loss', factor=0.1, patience=15,
                          verbose=1, mode='min', min_delta=0.0001, cooldown=1,
                          min_lr=0.00001)
cb_es = EarlyStopping(monitor="val_loss", patience=20, mode="min", restore_best_weights=True)
cb_tb = TensorBoard(log_dir=os.path.join(fold_subdir, "tensorboard"),
                    histogram_freq=0, write_graph=True, write_images=True)
cb_cl = CSVLogger(os.path.join(fold_subdir, "logs.csv"), separator=',',
                  append=True)
cb_ckp = ModelCheckpoint(os.path.join(fold_subdir, "model.hdf5"),
                         save_best_only=True,
                         monitor="val_loss",
                         mode="min")

#-----------------------------------------------------#
#                           Train                            #
#-----------------------------------------------------#
random.shuffle(sample_list)

train_list = sample_list[:2000]
val_list = sample_list[2000:3000]
test_list = sample_list[3000:]

model.evaluate(train_list,
               val_list,
               epochs=1000,
               iterations=250,
               callbacks=[cb_lr, cb_es, cb_tb, cb_cl, cb_ckp])

## Dump model to disk for reproducibility
model.dump(os.path.join(fold_subdir, "model.hdf5"))

Thanks in advance!! Constanza

[covasquezv]

I solved my problem. I just needed to change the interface from: interface = NIFTI_interface(channels=1, classes=2) to interface = NIFTI_interface(channels=1, classes=2, three_dim=False).

Now it works perfectly! I leave it here in case anyone is facing the same issue.