If there any details about how to fine-tune the resnet101
NanAlbert opened this issue · comments
This work is very interesting. But the code doesn't include the training of the feature extraction network. Can you provide corresponding code or training details? Thank you.
Hello @NanAlbert,
Regarding fine-tuning, what we mean is that the Imagenet-trained Resnet backbone is independently finetuned on the seen classes for each dataset. Once the finetuning is completed for the backbone, features for the images are extracted and used as input to the VAEGAN. This was the procedure used in the baseline f-VAEGAN-d2 paper (CVPR 19). The entire Resnet was finetuned with a low lr of 1e-5 or 1e-6.