search

vickyFox / Region-wise-Inpainting

Region-wise GAN, image inpainting
MIT License
53 stars 17 forks source link

Results #12

Open hhheeexxxuuu opened 3 years ago

hhheeexxxuuu commented 3 years ago

Hello, I have two questions about your results in the table2 and table3 of the paper. When you compare with other methods on the CelebA-HQ dataset, are all algorithms tested with 2000 images? What's more, I noticed that the mask has a range of values such as 0-10%. So did you take the intermediate value such as 5%, or take several values such as 1%, 5% and 10%, and then average the psnr results?

shihaobai commented 3 years ago

Thank for your attention! All algorithms were tested with 2000 images. Moreover, mask with a range of values such as 0-10% mean that the 2000 masks‘ area are randomly between 0-10%.

hhheeexxxuuu commented 3 years ago

Thank you!

hhheeexxxuuu commented 3 years ago

By the way, how do you set the parameters in the mask_online.py file to control the mask area between 0%-%10 or 10%-20% ?