the small data size training.

[Usernamezhx]

Thanks for your work. I trained the musicnn as your suggestion because I only have 1K train data. I finetune the musicnn last ten layer with the pretrained model you provided which is trained with MSD dataset. the result show me that:

Does that look right ? thanks in advace.

[minzwon]

It looks like your model is overfitted to your training data. I recommend you to fine-tune only the last 1~2 dense layers and also try strong data augmentation for your inputs. You can find a useful data augmentation tool here.

https://github.com/Spijkervet/torchaudio-augmentations.git

[Usernamezhx]

It looks like your model is overfitted to your training data. I recommend you to fine-tune only the last 1~2 dense layers and also try strong data augmentation for your inputs. You can find a useful data augmentation tool here.

https://github.com/Spijkervet/torchaudio-augmentations.git

thanks for your reply. I will try it.

[Usernamezhx]

I only try last dense layer.maybe it works. but the pr and roc is low than the model trained all layer.