DirectoryDataLoader

class directory_data_loader.DirectoryDataLoader(target_emotion_map=None, datapath=None, validation_split=0.2, out_channels=1, time_delay=None)[source]

DataLoader subclass loads image and label data from directory.

Parameters:
  • target_emotion_map – Optional dict of target emotion label values/strings and their corresponding label vector index values.
  • datapath – Location of image dataset.
  • validation_split – Float percentage of data to use as validation set.
  • out_channels – Number of image channels.
  • time_delay – Number of images to load from each time series sample. Parameter must be provided to load time series data and unspecified if using static image data.
load_data()[source]

Loads image and label data from specified directory path.

Returns:Dataset object containing image and label data.