See-in-the-Extremely-Dark (SIED)
收藏资源简介:
See-in-the-Extremely-Dark (SIED)数据集是为了解决极低光照环境下RAW图像增强的问题而创建的。数据集由三个精确的亮度级别(0.01-0.1 lux,0.001-0.01 lux,0.0001-0.001 lux)的极低光照RAW图像和高品质的sRGB参考图像组成。该数据集的创建过程包括在专业光学实验室中收集合格的低光照RAW图像,并通过调整相机参数捕获配对的低光照和正常光照图像。数据集还通过添加校准的噪声模型来模拟现实中的黑暗环境。SIED数据集为极低光照RAW图像增强方法提供了基准,并推动了低光照图像增强领域的发展。
The See-in-the-Extremely-Dark (SIED) dataset was created to address the problem of RAW image enhancement in extremely low-light environments. The dataset consists of low-light RAW images and high-quality sRGB reference images across three precise brightness levels: 0.01–0.1 lux, 0.001–0.01 lux, and 0.0001–0.001 lux. The dataset creation process involved collecting qualified low-light RAW images in a professional optical laboratory, and capturing paired low-light and normal-light images by adjusting camera parameters. It also simulates real-world dark environments by adding calibrated noise models. The SIED dataset provides a benchmark for extremely low-light RAW image enhancement methods, and promotes the development of the low-light image enhancement field.
SIED数据集概述
基本信息
- 数据集名称:SIED (Learning to See in the Extremely Dark)
- 相关论文:Learning to See in the Extremely Dark
- 作者单位:四川大学、电子科技大学、上海交通大学、超高清视频技术国家创新中心
数据集内容
- 数据集状态:原始训练和评估数据集即将发布 (Coming soon!)
- 包含子数据集:
- SIED数据集
- SID数据集
相关资源
- 预训练模型:
- 可通过Google Drive和百度云下载(提取码未提供)
使用方法
- 训练:
- 修改
datasets/dataset.py以适应环境 - 运行
python train.py
- 修改
- 测试:
- 运行
python evaluate.py
- 运行
可视化结果
- 包含Canon和Sony设备的视觉对比图
致谢
- 部分代码基于以下工作:
- WeatherDiff
- MIMO-UNet




