REAL (Raw and Event Acquired in Low-light)
收藏资源简介:
REAL数据集由华中科技大学与新加坡国立大学联合构建,是首个提供极低光照条件下(0.001-0.8勒克斯)像素级对齐的RAW图像、事件流及sRGB真值的三模态基准数据集。该数据集包含47,800组数据样本,通过同轴成像系统采集,完整保留了光子匮乏环境下的传感器噪声特征与运动纹理信息。数据创建过程严格模拟了动态场景下的短曝光条件,通过同步触发事件相机和RAW相机获取跨模态配对数据。该数据集旨在推动极端暗光环境下的事件-图像融合算法研究,为解决运动模糊、纹理丢失和噪声干扰等低光成像核心问题提供关键实验平台。
The REAL dataset, jointly constructed by Huazhong University of Science and Technology and the National University of Singapore, is the first three-modal benchmark dataset offering pixel-aligned RAW images, event streams and sRGB ground truths under extremely low-light conditions (0.001–0.8 lux). It consists of 47,800 data samples collected using a coaxial imaging system, and fully retains the sensor noise characteristics and motion texture information in photon-starved environments. The dataset’s creation process strictly simulates short-exposure conditions in dynamic scenes, and acquires cross-modal paired data by synchronously triggering event cameras and RAW cameras. This dataset aims to advance research on event-image fusion algorithms in extremely dark environments, providing a critical experimental platform for addressing core low-light imaging issues such as motion blur, texture loss and noise interference.
NEC-Diff 数据集概述
数据集基本信息
- 数据集名称:NEC-Diff
- 全称:Noise-robust event–RAW complementary diffusion for seeing motion in extreme darkness
- 发布状态:即将发布(Coming soon)
数据集来源与背景
- 相关论文:NEC-Diff: Noise-robust event–RAW complementary diffusion for seeing motion in extreme darkness
- 论文作者:Haoyue Liu, Jinghan Xu, Luxin Feng, Hanyu Zhou, Haozhi Zhao, Yi Chang, Luxin Yan
- 发表会议:CVPR 2026(已录用)
数据集内容与用途
- 核心目标:用于在极端黑暗环境下观测运动
- 技术基础:基于噪声鲁棒的事件-RAW互补扩散方法
数据获取与代码
- 数据集:即将发布
- 相关代码:即将发布
备注
- 本概述基于数据集详情页面提供的公开信息。
- 具体数据集内容、规模、格式及获取方式需等待官方正式发布。




