How does the V2Model add sharpening? #84

Open
opened 2026-01-29 21:41:27 +00:00 by claunia · 1 comment
Owner

Originally created by @xuewengeophysics on GitHub (Oct 12, 2021).

Personally, I think V2Model is better=-=
By saying V1 model looks more natural, did you mean that the eyes in V2Model are too white?

As V2Model also adds sharpening, so it may look sharper.

Skin looks much smoother in V2, V1 gives out better texture.
That's the major difference I can see.
I think you reduced/ optimized input noises in V2, causing this.

Originally posted by @shubhamcodez in https://github.com/TencentARC/GFPGAN/issues/48#issuecomment-907578170

Originally created by @xuewengeophysics on GitHub (Oct 12, 2021). > Personally, I think V2Model is better=-= > By saying V1 model looks more natural, did you mean that the eyes in V2Model are too white? > > As V2Model also adds sharpening, so it may look sharper. Skin looks much smoother in V2, V1 gives out better texture. That's the major difference I can see. I think you reduced/ optimized input noises in V2, causing this. _Originally posted by @shubhamcodez in https://github.com/TencentARC/GFPGAN/issues/48#issuecomment-907578170_
Author
Owner

@xuewengeophysics commented on GitHub (Oct 12, 2021):

Personally, I think V2Model is better=-=
By saying V1 model looks more natural, did you mean that the eyes in V2Model are too white?
As V2Model also adds sharpening, so it may look sharper.

Skin looks much smoother in V2, V1 gives out better texture. That's the major difference I can see. I think you reduced/ optimized input noises in V2, causing this.

Originally posted by @shubhamcodez in #48 (comment)

@xinntao
Hi xinntao,
Thanks for your excellent research work.
How does the V2Model add sharpening?
It depends on the noise?
Or depends on your training data?

@xuewengeophysics commented on GitHub (Oct 12, 2021): > > Personally, I think V2Model is better=-= > > By saying V1 model looks more natural, did you mean that the eyes in V2Model are too white? > > As V2Model also adds sharpening, so it may look sharper. > > Skin looks much smoother in V2, V1 gives out better texture. That's the major difference I can see. I think you reduced/ optimized input noises in V2, causing this. > > _Originally posted by @shubhamcodez in [#48 (comment)](https://github.com/TencentARC/GFPGAN/issues/48#issuecomment-907578170)_ @xinntao Hi xinntao, Thanks for your excellent research work. How does the V2Model add sharpening? It depends on the noise? Or depends on your training data?
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: TencentARC/GFPGAN#84