search

mmasana / FACIL

Framework for Analysis of Class-Incremental Learning with 12 state-of-the-art methods and 3 baselines.
https://arxiv.org/pdf/2010.15277.pdf
MIT License
512 stars 98 forks source link

The extensibility of the FACIL framework. #2

Closed Veagau closed 3 years ago

Veagau commented 3 years ago

Is it suitable for the more strict tasks, like the Few-shot Class Incremental Learning (FSCIL) scenario?

mmasana commented 3 years ago

Hi @Veagau! I do not understand what you mean by more strict tasks. Could you clarify what you mean? However, the framework could be adapted to do few-shot learning. I think the best way to start would be to make a copy of src/approach/incremental_learning.py and modify the train_loop() and train_epoch() functions to those of a few-shot learning setting. If necessary, then some changes could be made to the main_incremental.py (below line 221). If you define a bit more the specifics of what you would like to try, I might be able to provide more insight or code.