dcase_util.datasets.Dataset.validation_split
- Dataset.validation_split(fold=None, training_meta=None, split_type='balanced', validation_amount=None, seed=0, verbose=False, scene_label=None, iterations=100, **kwargs)[source]
List of validation files. Validation files are always subset of training files.
- Parameters
- foldint
Fold id, if None all meta data is returned. Default value None
- training_metaMetaDataContainer
Training data meta container. Use this instead of fold parameter, if additional processing is needed for training meta before usage. Default value None
- split_typestr
Split type [dataset, random, balanced] Default value ‘balanced’
- validation_amountfloat
Amount of training files to be assigned for validation Default value None
- seedint
Randomization seed Default value 0
- verbosebool
Show information about the validation set. Default value False
- scene_labelstr
Scene label of the validation set. If None, all training material used. Default value None
- iterationsint
Randomization iterations done when finding balanced set before selecting best matched set. Default value 100
- Returns
- list of str
List containing all files assigned to training set for given fold.
- list of str
List containing all files assigned to validation set for given fold.