Datasets

Dataset loading utilities for training and evaluation.

Image Datasets

ImageFolder

Custom ImageFolder dataset for training compression models.

ImageFolder

ImageFolder(root, transform=None, split='train')

Bases: Dataset[Image | T]

Load an image folder database. Training and testing image samples are respectively stored in separate directories:

.. code-block::

- rootdir/
    - train/
        - img000.png
        - img001.png
    - test/
        - img000.png
        - img001.png

Parameters:

Name	Type	Description	Default
root	string	root directory of the dataset	required
transform	callable	a function or transform that takes in a PIL image and returns a transformed version	None
split	string	split mode ('train' or 'val')	'train'

__getitem__

__getitem__(index)

Parameters:

Name	Type	Description	Default
index	int	Index	required

Returns:

Name	Type	Description
img	Image | T	PIL.Image.Image or transformed PIL.Image.Image.

Video Datasets

VideoFolder

VideoFolder(root, rnd_interval=False, rnd_temp_order=False, transform=None, split='train')

Bases: Dataset[tuple[Tensor, ...]]

Load a video folder database. Training and testing video clips are stored in a directorie containing mnay sub-directorie like Vimeo90K Dataset:

.. code-block::

- rootdir/
    train.list
    test.list
    - sequences/
        - 00010/
            ...
            -0932/
            -0933/
            ...
        - 00011/
            ...
        - 00012/
            ...

training and testing (valid) clips are withdrew from sub-directory navigated by corresponding input files listing relevant folders.

This class returns a set of three video frames in a tuple. Random interval can be applied to if subfolders includes more than 6 frames.

Parameters:

Name	Type	Description	Default
root	string	root directory of the dataset	required
rnd_interval	bool	enable random interval [1,2,3] when drawing sample frames	False
transform	callable	a function or transform that takes in a PIL image and returns a transformed version	None
split	string	split mode ('train' or 'test')	'train'

__getitem__

__getitem__(index)

Parameters:

Name	Type	Description	Default
index	int	Index	required

Returns:

Name	Type	Description
img	tuple[Tensor, ...]	PIL.Image.Image or transformed PIL.Image.Image.

Dataset Structure

Tinify datasets follow this structure:

dataset/
├── train/
│   ├── image001.png
│   ├── image002.png
│   └── ...
└── test/
    ├── image001.png
    ├── image002.png
    └── ...

Images should be in PNG or JPEG format with RGB channels.