I have a question relating to your FID Analysis:
According to the table you posted in the README, how exactly did you conduct the FID analysis ? Are you only using the 64x64 image output to compare with COCO validation set ? Or do you upsample it up to 256 (or 1024) and compare ? (In that case which SR model were you using ?)
Hi DeepFloyd Team,
Thanks for open-source your work!
I have a question relating to your FID Analysis: According to the table you posted in the README, how exactly did you conduct the FID analysis ? Are you only using the 64x64 image output to compare with COCO validation set ? Or do you upsample it up to 256 (or 1024) and compare ? (In that case which SR model were you using ?)