About ImageNet-A Benchmark Result.

[erjui]

Hi, thank you for the nice work first of all.

I wonder if it is possible to share the model's results on ImageNet-A?
I think this benchmark is widely used in domain generalization evaluations in this line of researches as well.

Thanks!

[Martinser]

Thank you for your interest in our work. Unfortunately, we have not yet conducted experiments on ImageNet-A. However, if you are interested in evaluating the performance of the model on ImageNet-A, you can do so by modifying the dataset path within the script. Specifically, you can add the ImageNet-A dataset path in the corresponding section of the train_cross_dataset.py file located in train_script/caspl_promptsrc/.

Feel free to run the script after making these adjustments, and it should allow you to test the model on ImageNet-A.

Should you encounter any issues or have further questions, please do not hesitate to reach out.

[erjui]

Thank you so much for the response!

Just minor request:
I tried to download the checkpoint to evaluate on the ImageNet-A benchmark, but I'm having little trouble with dealing with Baidu Network Disk (it requires me to sign up and download the client program, and it is little bit annoying to proceed).
So could I get the checkpoint for evaluating CasPL on ImageNet-A personally, or could you upload the checkpoint to other clouds (e.g., Google Drive) as well?

Thanks!