Hi,
First of all, many congrats for your ECCV-2020 paper. This method looks to be very effective on large-scale datasets like Imagenet.
We are trying to replicate your result on Imagenet dataset, for our research purpose.
In this repo, you have provided code for training on cifar10/100 dataset. However, for Imaganet, the network architecture changes, as well as training hyperparams.
As we want to replicate the results you got, without worrying about much changes in code, it would be great if you can provide ready-to-run code for imagenet experiment also(Results of your method for Resnet34-Resnet18 pair in Table 5).
Hi, First of all, many congrats for your ECCV-2020 paper. This method looks to be very effective on large-scale datasets like Imagenet.
We are trying to replicate your result on Imagenet dataset, for our research purpose. In this repo, you have provided code for training on cifar10/100 dataset. However, for Imaganet, the network architecture changes, as well as training hyperparams. As we want to replicate the results you got, without worrying about much changes in code, it would be great if you can provide ready-to-run code for imagenet experiment also(Results of your method for Resnet34-Resnet18 pair in Table 5).
Thank you