Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Image Restoration with Mean-Reverting Stochastic Differential Equations, ICML 2023. Winning solution of the NTIRE 2023 Image Shadow Removal Challenge.
| Date | Stars |
|---|---|
| 2026-07-24 | 720 |
| 2026-07-25 | 720 |
| 2026-07-28 | 720 |
| 2026-07-30 | 720 |
| 2026-08-06 | 720 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
## Image Restoration SDE <br><sub>Official PyTorch Implementations of [[IR-SDE]](https://proceedings.mlr.press/v202/luo23b.html)(ICML 2023) and [[Refusion]](https://arxiv.org/abs/2304.08291)(CVPRW 2023). </sub> [Ziwei Luo](https://algolzw.github.io/), [Fredrik K. Gustafsson](https://www.fregu856.com/), [Zheng Zhao](https://zz.zabemon.com/), [Jens Sjölund](https://jsjol.github.io/), [Thomas B. Schön](https://user.it.uu.se/~thosc112/index.html)</br> Department of Information Technology, Uppsala University</br> **[Image Restoration with Mean-Reverting Stochastic Differential Equations](https://proceedings.mlr.press/v202/luo23b.html)** [[Weights and Results](https://drive.google.com/drive/folders/14SvJXvp0HPKFHpGaaHy7h2I2fkRnIXAG?usp=share_link)]</br>  **[Refusion: Enabling Large-Size Realistic Image Restoration with Latent-Space Diffusion Models](https://arxiv.org/abs/2304.08291)** [[Weights and Results](https://drive.google.com/drive/folders/1Xy7z9CH0weck775ixePnjD6fZIjAbre7?usp=share_link)]</br>  Both methods can be used for synthetic and real-world datasets. You can find more details and results from this **[Project page](https://algolzw.github.io/ir-sde/index.html)**. ### Updates [**2024.04.15**] Updated a [posterior sampling](https://github.com/Algolzw/image-restoration-sde/blob/8c03a712f81f30db72338ff8a8000ae89ce29f68/codes/utils/sde_utils.py#L284) approach for better image generation. See paper [Photo-Realistic Image Restoration in the Wild with Controlled Vision-Language Models](https://arxiv.org/abs/2404.09732) for more details. <br> [**2023.10.02**] We present a new *Universal Image Restoration* framework by controlling vision-language models [[here]](https://github.com/Algolzw/daclip-uir).<br> [**2023.07.06**] All results and pre-trained models for IR-SDE and Refusion are provided in [link1](https://drive.google.com/drive/folders/14SvJXvp0HPKFHpGaaHy7h2I2fkRnIXAG?usp=share_link) and [link2](https://drive.google.com/drive/folders/1T5_UL4BgOIftEEkAzoa-SYX49P7COVa6?usp=share_link), respectively.<br> [**2023.04.25**] 🎉🎉🎉 **Our [IR-SDE](https://proceedings.mlr.press/v202/luo23b.html) paper was accepted by ICML 2023** 🎉🎉🎉<br> [**2023.04.23**] We released the codes for various tasks (including *IR-SDE*, *Refusion*, and *Latent-Refusion*).<br> [**2023.04.17**] 🎉🎉🎉 Our method **Refusion** is the **Winning Solution** in NTIRE 2023 Image Shadow Removal Challenge [[Paper]](https://arxiv.org/abs/2304.08291).<br> <!--[**2023.02.02**] We provide a [Gradio](https://gradio.app/) app file for the case that you want to [test your own images](https://github.com/Algolzw/image-restoration-sde#Gradio).--> ## Dependenices * OS: Ubuntu 20.04 * nvidia : - cuda: 11.7 - cudnn: 8.5.0 * python3 * pytorch >= 1.13.0 * Python packages: `pip install -r requirements.txt` ## How to use our Code? Here we provide an example for **image deraining** task, but can be changed to any problem with replacing the dataset. And the code for Denoising-SDE/ODE will be released later! We retrained the deraining model from scratch on 4 Nvidia TITAN XP GPUs and find that it achieves a **new SOTA even in terms of PSNR** on Rain100H: | Method | PSNR | SSIM | LPIPS | FID | | :--- | :----: | :----: | :----: | :----: | | **IR-SDE** | **31.65** | **0.9041** | **0.047** | **18.64** | | Restormer | 31.46 | 0.904 | - | - | | MPRNet | 30.41 | 0.8906 | 0.158 | 61.59 | | PReNet | 29.46 | 0.8990 | 0.128 | 52.67 | Note that **we didn't tune any parameter**, the last saved checkpoint was used to evaluation. The pretrained models for all tasks are provided [here](https://drive.google.com/drive/folders/14SvJXvp0HPKFHpGaaHy7h2I2fkRnIXAG?usp=share_link), and the performances of other SOTAs can be find in [here](https://paperswithcode.com/sota/single-image-deraining-on-rain100h). ### For Refusion Model Note that if you want to train/test the Refusion model, the only change is to replace the "**ir-s
Excerpt of 7,827 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:99cb086412d53f4e, topic:diffusion-models, topic:inpainting, topic:super-resolution
