Question about Multi-Class Classification Continual Learning in Avalanche

[karanwxliaa]

I have an image dataset where each image can belong to multiple classes, and I'm interested in training the data in a single experience/task with all the images and classes simultaneously.

Specifics:

• Dataset Structure: My dataset is structured in a way that one image can have multiple associated classes.
• Training Goal: I aim to train the data in a single experience/task, incorporating all images and their associated classes.

Questions:

1. Is it possible to use Avalanche for multi-class classification in a continual learning setting where each image can belong to multiple classes?
2. If yes, how would we incorporate this into a CL scenario and how can we form the dataloaders?