How to load a custom dataset (COCO format) for instance segmentation?

[jeanpat]

Hello,
I have a dataset available on kaggle consisting in 126 images and their segmentation labels in COCO format.

How can I modify the given example:

# 1. Create the DataModule
data_dir = icedata.pets.load_data()

datamodule = InstanceSegmentationData.from_folders(
    train_folder=data_dir,
    val_split=0.1,
    parser=partial(icedata.pets.parser, mask=True),
)

to load it?
Would it be possible then to introduce simply data augmentation (translations..., rotations...)?

Thank you

[rohitgr7]

hey @jeanpat!

the ques is related to flash. Would you mind asking it here instead: https://github.com/PyTorchLightning/lightning-flash/discussions?