|
|
--- |
|
|
library_name: pytorch |
|
|
license: other |
|
|
tags: |
|
|
- low-level-vision |
|
|
- all-in-one image-restoration |
|
|
language: |
|
|
- en |
|
|
pipeline_tag: image-to-image |
|
|
model-index: |
|
|
- name: RAM / RAM++ |
|
|
results: |
|
|
- task: |
|
|
type: image-to-image |
|
|
name: All-in-One Image Restoration |
|
|
dataset: |
|
|
name: placeholder |
|
|
type: image |
|
|
metrics: |
|
|
- name: PSNR |
|
|
type: psnr |
|
|
value: 0.0 |
|
|
--- |
|
|
This is the official pretrained models for the paper. |
|
|
>**Restore Anything with Masks：Leveraging Mask Image Modeling for Blind All-in-One Image Restoration**<br> [Chujie Qin](https://github.com/Dragonisss), [Ruiqi Wu](https://rq-wu.github.io/), [Zikun Liu](), [Xin Lin](https://linxin0.github.io/), [Chunle Guo](https://scholar.google.com/citations?user=RZLYwR0AAAAJ&hl=en), [Hyun Hee Park](s), [Chongyi Li<sup>†</sup>](https://li-chongyi.github.io/)<br/> |
|
|
> ( † indicates corresponding author )<br/> |
|
|
> In ECCV 2024, \[[HomePage](https://rq-wu.github.io/projects/RAM/index.html)\], \[[Paper Link](https://arxiv.org/abs/2409.19403v1)\] |
|
|
|
|
|
> **RAM++: <u>R</u>obust Representation Learning via <u>A</u>daptive <u>M</u>ask for All-in-One Image Restoration**<br> |
|
|
> [Zilong Zhang<sup>*</sup>](https://github.com/Zilong-Zhang003), [Chujie Qin<sup>*</sup>](https://github.com/DragonisCV), [Chunle Guo](https://mmcheng.net/clguo/), [Yong Zhang](), [Chao Xue](), [Ming-Ming Cheng](https://mmcheng.net/cmm/), [Chongyi Li<sup>†</sup>](https://li-chongyi.github.io/)<br/> |
|
|
> (<sup>*</sup>indicates equal contribution; <sup>†</sup> indicates corresponding author)<br/> |
|
|
> arxiv preprint, \[[HomePage](https://zilong-zhang003.github.io/RAM2.0/)\], \[[Paper Link](https://arxiv.org/abs/2509.12039)\] |
|
|
|
|
|
|
|
|
# Model description |
|
|
## RAM |
|
|
This method is architecture-agnostic and can be trained with any model. \ |
|
|
Here we provide the pre-trained and fine-tuned weights for two representative models: <strong>[PromptIR](https://github.com/va1shn9v/PromptIR)</strong> and <strong>[SwinIR](https://github.com/JingyunLiang/SwinIR)</strong>. |
|
|
## RAM_plus |
|
|
<strong>AdaSAM</strong> is a ViT-based, pixel-level mask generator. It analyzes correlations between image tokens and applies masks to regions that are semantically and texturally rich. |
|
|
|
|
|
<strong>RestormerWoSkip</strong> is built on <strong>[Restormer](https://github.com/swz30/Restormer)</strong>; it differs by removing the long-range residual connections. |
|
|
|
|
|
<strong>RestormerRFR</strong> regularizes via an efficient feature-fusion strategy that leverages DINOv2’s semantic consistency and degradation invariance. |
|
|
|
|
|
<strong>Different folders</strong> contain model weights trained under configurations with different numbers of tasks. |
|
|
|
|
|
# How to use |
|
|
For full instructions and runnable scripts, see the [code repository](https://github.com/DragonisCV/RAM/) |
|
|
## RAM |
|
|
### Pre-training： |
|
|
```python |
|
|
mask, mask_token = Random(img) #pixel_level |
|
|
output = PromptIR(img, mask, mask_token) |
|
|
``` |
|
|
### Fine-tuning: |
|
|
```python |
|
|
output = PromptIR(img, mask=None, mask_token=None) |
|
|
``` |
|
|
## RAM_plus |
|
|
### Pre-training： |
|
|
```python |
|
|
mask, mask_token = AdaSAM(img) |
|
|
output = RestormerWoSkip(img, mask, mask_token) |
|
|
``` |
|
|
### Fine-tuning: |
|
|
```python |
|
|
dino_features = DINOv2(img) |
|
|
output = RestormerRFR(img, mask=None, mask_token=None, dino_features) |
|
|
``` |
|
|
|
|
|
# Citation |
|
|
If you find our repo useful for your research, please consider citing our paper: |
|
|
```bibtex |
|
|
@inproceedings{qin2024restore, |
|
|
title={Restore Anything with Masks: Leveraging Mask Image Modeling for Blind All-in-One Image Restoration}, |
|
|
author={Qin, Chu-Jie and Wu, Rui-Qi and Liu, Zikun and Lin, Xin and Guo, Chun-Le and Park, Hyun Hee and Li, Chongyi}, |
|
|
booktitle={European Conference on Computer Vision}, |
|
|
pages={364--380}, |
|
|
year={2024}, |
|
|
organization={Springer} |
|
|
} |
|
|
|
|
|
@misc{zhang2025ramrobustrepresentationlearning, |
|
|
title={RAM++: Robust Representation Learning via Adaptive Mask for All-in-One Image Restoration}, |
|
|
author={Zilong Zhang and Chujie Qin and Chunle Guo and Yong Zhang and Chao Xue and Ming-Ming Cheng and Chongyi Li}, |
|
|
year={2025}, |
|
|
eprint={2509.12039}, |
|
|
archivePrefix={arXiv}, |
|
|
primaryClass={cs.CV}, |
|
|
url={https://arxiv.org/abs/2509.12039}, |
|
|
} |
|
|
``` |
|
|
|
