VGG-Trained-on-SIN Link not working

rgeirhos / texture-vs-shape

Pre-trained models, data, code & materials from the paper "ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness" (ICLR 2019 Oral)

https://openreview.net/forum?id=Bygh9j09KX

Other

785 stars 101 forks source link

VGG-Trained-on-SIN Link not working #21

Closed MKowal2 closed 4 years ago

MKowal2 commented 4 years ago

Not sure why, but the bitbucket link for the VGG16 network trained on SIN appears to be broken. Please let me know if this gets fixed!

https://bitbucket.org/robert_geirhos/texture-vs-shape-pretrained-models/raw/0008049cd10f74a944c6d5e90d4639927f8620ae/vgg16_train_60_epochs_lr0.01-6c6fcc9f.pth.tar

rgeirhos commented 4 years ago

I think https://github.com/rgeirhos/texture-vs-shape/blob/master/models/load_pretrained_models.py#L33 is what you're looking for.

MKowal2 commented 4 years ago

Perfect. Thanks for the quick reply!