[ICCV 2025] Learning to See in the Extremely Dark [Paper]

Hai Jiang¹, Binhao Guan², Zhen Liu², Xiaohong Liu³, Jian Yu⁴, Zheng Liu⁴, Songchen Han¹, Shuaicheng Liu²

1.Sichuan University,

2.University of Electronic Science and Technology of China,

3.Shanghai Jiaotong University,

4.National Innovation Center for UHD Video Technology

Dataset synthesis pipeline

Framework pipeline

Dependencies

pip install -r requirements.txt

Download the raw training and evaluation datasets

SIED dataset

Our SIED dataset is available at [OneDrive] and [Baidu Yun (extracted code:4y4w)]. Please see the txt files in data folder for the training set and evaluation set split.

SID dataset

Pre-trained Models

You can download our pre-trained model from [OneDrive] and [Baidu Yun (extracted code:m9zp)]

How to train?

You need to modify dataset/dataloader.py slightly for your environment, and then

accelerate launch train.py  

How to test?

python inference.py

Visual comparison

Citation

If you use this code or ideas from the paper for your research, please cite our paper:

@inproceedings{sied,
    author    = {Jiang, Hai and Guan, Binhao and Liu, Zhen and Liu, Xiaohong and Yu, Jian and Liu, Zheng and Han, Songchen and Liu, Shuaicheng},
    title     = {Learning to See in the Extremely Dark},
    booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision},
    year      = {2025},
    pages     = {7676-7685}
}

Acknowledgement

Part of the code is adapted from the previous work: denoising-diffusion-pytorch. We thank all the authors for their contributions.