search

G-U-N / ECCV22-FOSTER

The official implementation for ECCV22 paper: "FOSTER: Feature Boosting and Compression for Class-Incremental Learning" in PyTorch.
MIT License
51 stars 4 forks source link

The CNN top1 curve in experiment #6

Closed zhl98 closed 2 years ago

zhl98 commented 2 years ago

Hi! First of all, thank you very much for your repo. I am very interested in your paper. I find the darknet eval is the top1 acc of the snet. However, its precision is not the same as that in the last list. Later, I found that the model used the original teacher network with two backbone networks when evaluating. Because you update the self._network is in the begin of incremental_train(). This is the original plan to use the result of the teacher model, or is it a mistake. image

G-U-N commented 2 years ago

Thank you very much for pointing this out, and you should use the accuracy in darknet eval as the result of the single-backbone model. If you hope to directly get those accuracy-lists of the single-backbone model, please add the following code

self._network = self._snet

in line 133, foster.py.

Hope this clarifies your queries.