deeptrack.sources.folder Module#

Utility class for data sources in a directory structure.

This module provies the ImageFolder DeepTrack2 class which enables control of image sources organized in a directory structure.

The primary usage is to facilitate naming and organizing of data sources.

Key Features#

  • Attribute Access

    Enables accessing attributes tied to a data source such as paths, directory structure, length etc.

  • Labeling

    Allows converting category names of images to integers, which is more flexible and easy to process in a data pipeline.

  • Category Splitting

    The sources of images can be split into subcategories of which the user specifies the name of.

Module Structure#

ImageFolder: Data source for images organized in a directory structure.

Allows for processing of image sources with Dict data strucutres, splitting, naming and labeling functions.

Examples#

Print some information about a source of data:

>>> from deeptrack.sources import folder

>>> root = "data/train"
>>> data_source = folder.ImageFolder(root)

>>> print(f"Total images in training data: {len(train_data)}")
>>> print(f"Classes: {train_data.classes}")

Classes#

ImageFolder(root)

Data source for images organized in a directory structure.

Source(**kwargs)

A class that represents one or more sources of data.