I'm new to mmsegmentation but quite familiar with PyTorch.
I would like to create a dataset class that can be used in mmsegmentation without modifying the dataset file structure.
From what I've found, it seems necessary to change the file structure and ensure all images have the same suffix.
However, I prefer not to do that. Instead, I want to write my own getitem method in a class inheriting from torch.utils.data.Dataset to easily customize the loading process.
Could you please provide guidance on how to achieve this or refer me to any relevant documentation?
Hi,
I'm new to mmsegmentation but quite familiar with PyTorch. I would like to create a dataset class that can be used in mmsegmentation without modifying the dataset file structure. From what I've found, it seems necessary to change the file structure and ensure all images have the same suffix. However, I prefer not to do that. Instead, I want to write my own getitem method in a class inheriting from torch.utils.data.Dataset to easily customize the loading process.
Could you please provide guidance on how to achieve this or refer me to any relevant documentation?
Thank you.