Skip to content

No prompt for ImageDescriptionGenerator #5186

Open
@dotMorten

Description

@dotMorten

The ImageDescriptionGenerator is missing an option to make a prompt to guide the response to what sort of description is requested. The prompt could be to ignore objects in the background, or ask specific questions about the subject in the image.

When I was using the Phi3Vision model directly this was a huge benefit and opened up quite a few scenarios for analyzing images.

Metadata

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions