MultiDiffusion

MultiDiffusion: Fusing Diffusion Paths for Controlled Image Generation

MultiDiffusion是一个统一的生成框架，它通过一个最小二乘优化任务，将同一个扩散模型生成的多个crop约束在一起，虽然最初各个crop生成的内容可能会不一致，但MultiDiffusion会通过全局去噪，不断融合各个crop，从而产生高质量和连贯性的图像。

example

if __name__ == "__main__":

    from modelscope.pipelines import pipeline

    model = 'jianguo_wjg/multidiffusion'
    pipe = pipeline('multidiffusion', model=model, model_revision="v0.1.6")

    H = 512
    W = 2048

    prompt = "a photo of the dolomites"
    image = pipe(prompt, height=H, width=W).images[0]
    image.save("dolomites.jpg")

License

The model is licensed under the MIT license.

Citation

@article{bar2023multidiffusion,
  title={MultiDiffusion: Fusing Diffusion Paths for Controlled Image Generation},
  author={Bar-Tal, Omer and Yariv, Lior and Lipman, Yaron and Dekel, Tali},
  journal={arXiv preprint arXiv:2302.08113},
  year={2023}
}

Clone with HTTP

 git clone https://www.modelscope.cn/jianguo_wjg/multidiffusion.git