SDE
收藏资源简介:
SDE数据集由香港科技大学(广州)信息枢纽人工智能学域创建,包含超过31,000对在低光和正常光照条件下捕获的时空对齐视频帧和事件。数据集通过机器人臂精确对齐,确保了空间误差小于0.03mm和时间误差小于0.01s。创建过程中,数据集还添加了伪语义分割和深度估计标签,以支持下游任务的评估。该数据集主要应用于低光视频增强,旨在解决在极端低光条件下图像质量下降的问题。
The SDE Dataset was developed by the AI Discipline of the Information Hub at The Hong Kong University of Science and Technology (Guangzhou). It contains over 31,000 pairs of spatially and temporally aligned video frames and event data captured under both low-light and standard illumination conditions. The dataset is precisely aligned using a robotic arm, guaranteeing a spatial error of less than 0.03 mm and a temporal error of less than 0.01 s. During the construction process, pseudo semantic segmentation and depth estimation labels were additionally added to facilitate evaluation for downstream tasks. This dataset is mainly utilized for low-light video enhancement, with the goal of resolving the problem of deteriorated image quality under extreme low-light conditions.




