Skip to content

phymhan/prompt-to-prompt

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

28 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Improving Tuning-Free Real Image Editing via Proximal Guidance

The code is heavily based on the Prompt-to-Prompt, Null-Text Inversion [codebase], the MasaCtrl [codebase], and the DDPM-Inversion [codebase].

Input images can be downloaded through this link. Most images are from the StyleDiffusion, Null-Text Inversion, Imagic, MasaCtrl, and SVDiff papers.

Examples

Negative-Prompt Inversion (NPI):

from negative_prompt_inversion import main
main(
    image_path="./images/cat_mirror.png",
    prompt_src="a cat sitting next to a mirror",
    prompt_tar="a silver cat sculpture sitting next to a mirror",
    output_dir="./outputs/npi_real",
    suffix="silver",
    guidance_scale=7.5,
    cross_replace_steps=0.8,
    self_replace_steps=0.6,
    blend_word=((('cat',), ("cat",))),
    eq_params={"words": ("silver", 'sculpture', ), "values": (2,2,)},
    proximal=None,
)

NPI reconstructed (left) | NPI edited (right):
cat-mirror-npi

To run ProxNPI:

from negative_prompt_inversion import main
main(
    image_path="./images/cat_mirror.png",
    prompt_src="a cat sitting next to a mirror",
    prompt_tar="a silver cat sculpture sitting next to a mirror",
    output_dir="./outputs/npi_real",
    suffix="silver",
    guidance_scale=7.5,
    cross_replace_steps=0.8,
    self_replace_steps=0.6,
    blend_word=((('cat',), ("cat",))),
    eq_params={"words": ("silver", 'sculpture', ), "values": (2,2,)},
    proximal='l0',
    quantile=0.7,
    use_inversion_guidance=True,
    recon_lr=1,
    recon_t=400,
)

Reconstructed (left) | Edited (right):
cat-mirror-proxnpi

To run MasaCtrl:

from prox_masactrl import main
main(
    out_dir="./outputs/masactrl_real/",
    source_image_path="./images/cake2.png",
    source_prompt="a round cake",
    target_prompt="a square cake",
    npi=False,
    npi_interp=0,
)

Input image (left) | Reconstructed (middle) | Edited (right):
cake-masa

To run ProxMasaCtrl:

from prox_masactrl import main
main(
    out_dir="./outputs/masactrl_real/",
    source_image_path="./images/cake2.png",
    source_prompt="a round cake",
    target_prompt="a square cake",
    npi=True,
    npi_interp=1,
    prox='l0',
    quantile=0.6,
)

Input image (left) | Reconstructed (middle) | Edited (right):
cake-proxmasa

Please see run_npi.py, run_masa.py and run_ddpm.py for examples.

Citation

@inproceedings{han2024proxedit,
  title={ProxEdit: Improving Tuning-Free Real Image Editing With Proximal Guidance},
  author={Han, Ligong and Wen, Song and Chen, Qi and Zhang, Zhixing and Song, Kunpeng and Ren, Mengwei and Gao, Ruijiang and Stathopoulos, Anastasis and He, Xiaoxiao and Chen, Yuxiao and others},
  booktitle={Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision},
  pages={4291--4301},
  year={2024}
}
@article{miyake2023negative,
  title={Negative-prompt Inversion: Fast Image Inversion for Editing with Text-guided Diffusion Models},
  author={Miyake, Daiki and Iohara, Akihiro and Saito, Yu and Tanaka, Toshiyuki},
  journal={arXiv preprint arXiv:2305.16807},
  year={2023}
}
@inproceedings{mokady2023null,
  title={Null-text inversion for editing real images using guided diffusion models},
  author={Mokady, Ron and Hertz, Amir and Aberman, Kfir and Pritch, Yael and Cohen-Or, Daniel},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={6038--6047},
  year={2023}
}
@article{hertz2022prompt,
  title={Prompt-to-prompt image editing with cross attention control},
  author={Hertz, Amir and Mokady, Ron and Tenenbaum, Jay and Aberman, Kfir and Pritch, Yael and Cohen-Or, Daniel},
  journal={arXiv preprint arXiv:2208.01626},
  year={2022}
}
@article{cao2023masactrl,
  title={MasaCtrl: Tuning-Free Mutual Self-Attention Control for Consistent Image Synthesis and Editing},
  author={Cao, Mingdeng and Wang, Xintao and Qi, Zhongang and Shan, Ying and Qie, Xiaohu and Zheng, Yinqiang},
  journal={arXiv preprint arXiv:2304.08465},
  year={2023}
}
@article{song2020denoising,
  title={Denoising diffusion implicit models},
  author={Song, Jiaming and Meng, Chenlin and Ermon, Stefano},
  journal={arXiv preprint arXiv:2010.02502},
  year={2020}
}
@article{HubermanSpiegelglas2023,
  title      = {An Edit Friendly DDPM Noise Space: Inversion and Manipulations},
  author     = {Huberman-Spiegelglas, Inbar and Kulikov, Vladimir and Michaeli, Tomer},
  journal    = {arXiv preprint arXiv:2304.06140},
  year       = {2023}
}

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 99.3%
  • Python 0.7%