In commit 25e6204d97ee2157360ee93a1a65d07523326b82, the way that CSV data files were handled was last modified. Specifically, in niftynet.io.image_sets_partitioner.ImageSetsPartitioner, method load_data_sections_by_subject:
for section_name in self.data_param:
if isinstance(self.data_param[section_name], dict):
mod_spec = self.data_param[section_name]
else:
mod_spec = vars(self.data_param[section_name])
if mod_spec.get('csv_data_file', None): # has csv_data_file
# skip file search
continue
This skips all sections where a 'csv_data_file' has been specified. I can't find the part of the code where these sections are supposed to be loaded, and can't find a reference documenting this.
I was looking into this because currently I have to perform a "hack" in order to input the class label as conditioning data in GANApplication, where I need to pass the label as a 1x1x1 image for each subject. Also, for ClassificationApplication, how would the class label be specified? Is this a bug? I am willing to submit a pull request fixing this, but I wanted to make sure that I understand the reason behind this fully.
In commit 25e6204d97ee2157360ee93a1a65d07523326b82, the way that CSV data files were handled was last modified. Specifically, in
niftynet.io.image_sets_partitioner.ImageSetsPartitioner
, methodload_data_sections_by_subject
:This skips all sections where a
'csv_data_file'
has been specified. I can't find the part of the code where these sections are supposed to be loaded, and can't find a reference documenting this.I was looking into this because currently I have to perform a "hack" in order to input the class label as conditioning data in
GANApplication
, where I need to pass the label as a 1x1x1 image for each subject. Also, forClassificationApplication
, how would the class label be specified? Is this a bug? I am willing to submit a pull request fixing this, but I wanted to make sure that I understand the reason behind this fully.