A problem in runtime

[wwlaoxi]

HI: I run the code of tensorflow-u-net and find a problem.I downloaded the traindata from LITS。 The traindata have 130 volume and 130 segmentation.The 0 to 95 on the traindata are the training sets. The 96 to 130 is a validation sets.when I run the code, there are two problem.I want to know the reason. 1、 The segmentation.py in the /dpp/twodim/segmentation robust_img_scaling(source, ignore_values=[], initialization_generator=None, initialize=True): when initialize=True ,it produces the following errors. If initialize=False,it is OK. so I want to known the reason.

Loading volume-6.nii Traceback (most recent call last): File "/home/PycharmProjects/tensorflow-u-net/training.py", line 525, in main() File "/home/PycharmProjects/tensorflow-u-net/training.py", line 83, in main validation_set = list(validation_pipeline) File "/home/PycharmProjects/tensorflow-u-net/dpp/helper.py", line 105, in next return self.current_generator.next() File "/home/PycharmProjects/tensorflow-u-net/dpp/helper.py", line 18, in _generator_func yield transform(datapoint) File "/home/PycharmProjects/tensorflow-u-net/dpp/twodim/segmentation/segmentation.py", line 386, in transformation scaler_params = dictionary[file_name] KeyError: 'volume-6.nii'

2、The liver_generate_predictions.py networks.parameter_efficient(in_channels=in_channels, out_channels=2, start_filters=90, input_side_length=256, sparse_labels=True, batch_size=batch_size, activation=activation_function, batch_norm=batch_norm) the start_filters=90 is wrong , the start_filters=64 is right

At last,can you send me some material or paper for the dpp (The Data Processing Pipeline library).for example ,the principle、 processing flow and so on.

Thanks long