Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

recommended parameters and image size for cleanface_HD #8

Open
tonytan48 opened this issue Apr 13, 2020 · 3 comments
Open

recommended parameters and image size for cleanface_HD #8

tonytan48 opened this issue Apr 13, 2020 · 3 comments
Labels
enhancement New feature or request todo I will solve it

Comments

@tonytan48
Copy link

Hi, I am wondering what is the recommended setting for this pretrained model. I have image of shape (1242, 2208, 3) and default setting, it returns can't find mosaic. I tuned --tr_blur and --tr_down parameter to 100 and still cannot find mosaic. For this image size, is it recommended to shrink the image first?

@HypoX64
Copy link
Owner

HypoX64 commented Apr 14, 2020

According to my guess, it may happend when the image is too large and mosaic is too small. When input images, the program may first reize the image to (224,224) to find where is the mosaic. In some times, it may loss mosaic. I will use new model to solve it.

@tonytan48
Copy link
Author

Thanks for the prompt reply. Since the images will be resized to squared shape, is it better to crop the image to square

@HypoX64 HypoX64 added enhancement New feature or request todo I will solve it labels Apr 19, 2020
HypoX64 added a commit that referenced this issue Apr 28, 2020
@BrandonC150
Copy link

Best way I've found to repair even severe mosaic is to use VirtualDub2 to "WarpResize" to 4times the origional resolution, "add mosaic", "Bilinear resize" to origional resolution, and finally clean mosaic! It yields an even blur rather than having very noticeable artifacts😸 and works for pretty much any mosaic

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
enhancement New feature or request todo I will solve it
Projects
None yet
Development

No branches or pull requests

3 participants