Question Regarding FID Scores #22
Description
Dear Author,
I have been carefully reviewing your paper, and I noticed that the FID scores reported are significantly lower than that in the original GANWriting paper, where the scores are typically above 80. I was wondering if you could kindly clarify whether the FID scores in your work are directly referenced from the original GANWriting paper, or if they reflect the results of your own reproduction of the experiments.
Additionally, I noticed that the paper does not provide detailed information on the exact methodology used for FID computation. This has made it challenging for me to fully understand how such low scores were achieved, as I was unable to replicate them under similar conditions.
Could you kindly shed some light on any potential differences in experimental setup, preprocessing, or evaluation criteria that might explain this discrepancy?
I greatly appreciate your time and any insights you can share.