Yes, it is true that there are many frameworks available for setting up datasets, such as ImageFolder, which is commonly used in computer vision tasks. However, in the real world, the data you may encounter may not always be in a format that can be directly used with these frameworks.

This would require you to build your own dataset and preprocess your data accordingly

With my code, you can:

• Use your dataset in your path folder
• You can customize your preprocessing steps based on the specific needs of your task

• Python
• Torch
• PIL
• Torchvision
• os

I have explain more clearly in my medium