Inst-Inpaint

Inst-Inpaint is a diffusion-based model that performs text-guided object removal from images. This model is a Cog wrapper implemented using the original repository.

Using the Model

To use the model, simply upload the image you want to edit and enter a text instruction to remove target object/s (e.g. “remove the can on the background”, “remove the kite”.)

Note that Inst-Inpaint is trained on and outputs 256 x 256 images, optionally chain Inst-Inpaint with super-resolution models to upscale generated images to higher resolution.

References

@misc{yildirim2023instinpaint,
      title={Inst-Inpaint: Instructing to Remove Objects with Diffusion Models}, 
      author={Ahmet Burak Yildirim and Vedat Baday and Erkut Erdem and Aykut Erdem and Aysegul Dundar},
      year={2023},
      eprint={2304.03246},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}