Skip to content

Lack of functionality in stable diffusion inpainting  #370

Closed
@Inkorak

Description

@Inkorak

The img2img can handle init_image as FloatTensor and PIL image with

if not isinstance(init_image, torch.FloatTensor):
   init_image = preprocess(init_image)

But inpainting cannot work like this with init_image and mask_image. It's only working with FloatTensor. Although the typing states that it can work with both.

init_image: Union[torch.FloatTensor, PIL.Image.Image],
mask_image: Union[torch.FloatTensor, PIL.Image.Image],

It is desirable to have uniformity.
Also, preprocess function in img2img and preprocess_image function in inpainting are the same, they differ only in the name.

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions