Dear authors,
thank you for releasing the code! I would like to use it to train domain classifiers to distinguish between arbitrary domains, e.g. the styles of different painters, such as has been the task in https://arxiv.org/pdf/2404.01292.pdf.
I think the most straight-forward way to do so would be to write the different domain datasets in ffcv format and write a custom ImageNetWithOODFFCVDataset where the indices go over the different styles. Am I understanding the code correctly or would you advise me to approach the task differently?
Best,
Evgenia
Dear authors,
thank you for releasing the code! I would like to use it to train domain classifiers to distinguish between arbitrary domains, e.g. the styles of different painters, such as has been the task in https://arxiv.org/pdf/2404.01292.pdf.
I think the most straight-forward way to do so would be to write the different domain datasets in ffcv format and write a custom ImageNetWithOODFFCVDataset where the indices go over the different styles. Am I understanding the code correctly or would you advise me to approach the task differently?
Best,
Evgenia