Add image-to-image pipeline tag and link to paper
This PR adds the pipeline_tag: image-to-image to the model card metadata. This will allow the model to be properly categorized and discovered on the Hugging Face Model Hub. Also adding a direct link to the paper.
Hello, thanks for the suggestions!
I think our model is not exactly an 'image-to-image' model, because its main input are a set of triangles but not an image. If we consider the view-representation token as the input image, then it is remotely an 'image-to-image' model, but this may make every diffusion image generation model an image(noised)-to-image model as well.
I think the ideal tag would be something like '3D-to-image', the inverse of 'image-to-3D', but currently HF does not have such a tag yet.
If you think having the 'image-to-image' category tag is still helpful, I can merge the PR (also in other variant models) to reflect them.