yuanzhi-zhu/DiffPIR
A diffusion model-based method for plug-and-play image restoration that uses pre-trained DDPM as a generative denoiser prior.
Not currently ranked — collecting fresh signals.
star history
This project applies denoising diffusion probabilistic models to solve various image restoration inverse problems. It leverages off-the-shelf diffusion models as implicit priors in plug-and-play frameworks, enabling training-free restoration across tasks like deblurring, inpainting, and super-resolution. The method builds on OpenAI guided diffusion and DPIR, using the diffusion model’s generative capabilities as a flexible denoiser.
Frequently asked
- What is yuanzhi-zhu/DiffPIR?
- A diffusion model-based method for plug-and-play image restoration that uses pre-trained DDPM as a generative denoiser prior.
- Is DiffPIR open source?
- Yes — yuanzhi-zhu/DiffPIR is open source, released under the MIT license.
- What language is DiffPIR written in?
- yuanzhi-zhu/DiffPIR is primarily written in Python.
- How popular is DiffPIR?
- yuanzhi-zhu/DiffPIR has 496 stars on GitHub.
- Where can I find DiffPIR?
- yuanzhi-zhu/DiffPIR is on GitHub at https://github.com/yuanzhi-zhu/DiffPIR.