Skip to content

Anime content training #41

Description

@vicentecarro

Hi,
Regarding training for Anime content, which differs significantly from real world content in terms of surface normals (due to stylization, shading conventions, and art stuff), from your experience, would it make more sense to (re)train the model from scratch, or to fine-tune the existing one, despite the domain gap?
If the latter, do you have any recommendations? Actually, in both cases, do you have any recommendations in general?

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions