trw.train.sample_export
¶
We assume throughout that the image format is (samples, channels, height, width)
Module Contents¶
Functions¶
|
Export a value as an image |
|
|
|
|
Default export functions |
|
|
|
|
Export a sample from the batch |
- trw.train.sample_export.export_as_image(name, samples, sample_id, export_root, txt_file)¶
Export a value as an image :param name: :param samples: :param export_root: :param txt_file: :return:
- trw.train.sample_export.export_as_npy(name, samples, sample_id, export_root, txt_file)¶
- trw.train.sample_export.export_as_string(name, samples, sample_id, export_root, txt_file, max_len=1024)¶
- trw.train.sample_export.export_functions()¶
Default export functions :return:
- trw.train.sample_export.clean_mapping_name(name)¶
- trw.train.sample_export.export_sample(batch, sample_id, export_root, txt_file, exports=export_functions, features_to_discard=None, clean_mapping_name_fn=clean_mapping_name, classification_mappings=None)¶
Export a sample from the batch :param batch: a batch :param sample_id: the index of the sample to export :param export_root: where to export the data (typically including the sample id) :param txt_file: where to export text data. Must be an opened file for write :param exports: a functor returning the functions to be used for the export :param features_to_discard: a list of feature names to discard :param clean_mapping_name_fn: function to translate the name of batch feature name to class output name :param classification_mappings: a dictionary of mapping output to translate class ID to class name