Thanks for your sharing your code. It's a wonderful job I think~~
I have one question about the content loss score. I have applied StyTr2 to a dataset of 800 images, using your pre-trained model. To ensure consistency with your test settings, I resized all the images to 256x256 before calculating the content loss. However, I have noticed significant differences in the content loss values compared to what is reported in your papers.
I understand that variations in scores are expected due to the use of different images. Nonetheless, I found that the style loss scores exhibit a similar trend, while the content loss scores demonstrate noticeable discrepancies. So May I know how can you calculate the content loss? Is it possible to share your metric code or tell me where I am wrong?
Thanks for your sharing your code. It's a wonderful job I think~~
I have one question about the content loss score. I have applied StyTr2 to a dataset of 800 images, using your pre-trained model. To ensure consistency with your test settings, I resized all the images to 256x256 before calculating the content loss. However, I have noticed significant differences in the content loss values compared to what is reported in your papers.
I understand that variations in scores are expected due to the use of different images. Nonetheless, I found that the style loss scores exhibit a similar trend, while the content loss scores demonstrate noticeable discrepancies. So May I know how can you calculate the content loss? Is it possible to share your metric code or tell me where I am wrong?