target must be b, c, x, y(, z) with c=1 :param net_output: :param target: :return:

Uses one binary channel: 1 - fg, 0 - bg pred: (b, 1, x, y, z) or (b, 1, x, y) target: (b, 1, x, y, z) or (b, 1, x, y)

Uses one binary channel: 1 - fg, 0 - bg pred: (b, 1, x, y, z) or (b, 1, x, y) target: (b, 1, x, y, z) or (b, 1, x, y)

net_output: (batch_size, class, x,y,z) target: ground truth, shape: (batch_size, 1, x,y,z) bound: precomputed distance map

net_output: (batch_size, 2, x,y,z) target: ground truth, shape: (batch_size, 1, x,y,z)

For prediction of test cases just set training=False, this will prevent loading of training data and training batchgenerator initiali

we need to swap out get_default_augmentation so that we can move the EDTs around (get_default_augmentation_withEDT)

For prediction of test cases just set training=False, this will prevent loading of training data and training batchgenerator initiali

customized nll loss source: https://medium.com/@zhang_yang/understanding-cross-entropy- implementation-in-pytorch-softmax-log-softmax-nll

customized nll loss source: https://medium.com/@zhang_yang/understanding-cross-entropy- implementation-in-pytorch-softmax-log-softmax-

Use this to predict new data :param input_files: :param output_file: :param softmax_ouput_file: :return:

here we add the boundary to the loss function :param data_generator: :param do_backprop: :param run_online_evaluatio

:param do_mirroring: :param use_train_mode: :param mirror_axes: :param tiled: :param tile_in_z: :par

2018_12_05: I added global accumulation of TP, FP and FN for the validation in here. This is because I believe that selecting models