[Feature] Add Albumentation capabilities to the dataset pipeline

[gvalvano]

What's the feature?

In MMDetection, it is possible to use Albumentations to make complex transformations to the input data (reference here and here).

Since fine-tuning generative models on small datasets could be problematic (e.g. dreambooth), I believe it would be extra-useful to add albumentation support in MMagic, too! This way, the user could artificially generate new data samples of the target concept.

Any other context?

Example of config file with albumentation:

albu_transforms = [
    dict(
        type='ShiftScaleRotate',
        shift_limit=0.0625,
        scale_limit=0.0,
        rotate_limit=0,
        interpolation=1,
        p=0.5),
    dict(
        type='RandomBrightnessContrast',
        brightness_limit=[0.1, 0.3],
        contrast_limit=[0.1, 0.3],
        p=0.2),),
]
train_pipeline = [
    dict(type='LoadImageFromFile', key='img', channel_order='rgb'),
    dict(type='Resize', scale=(512, 512)),
    dict(type='Albu', transforms=albu_transforms),
    dict(type='PackInputs')
]

[Z-Fran]

We will check it

[Z-Fran]

we will support a simple Albumentation there. #1894