train_tools.data_utils.utils

Module Contents

train_tools.data_utils.utils.split_train_valid(data_dicts, valid_portion=0.1)

Split train/validata data according to the given proportion

train_tools.data_utils.utils.path_decoder(root, mapping_file, no_label=False, unlabeled=False)

Decode img/label file paths from root & mapping directory.

Parameters:

• root (str)

• mapping_file (str) – json file containing image & label file paths.

• no_label (bool, optional) – whether to include “label” key. Defaults to False.

Returns:

list of dictionary. (ex. [{“img”: img_path, “label”: label_path}, …])

Return type:

list