DeepHDRVideo
收藏资源简介:
DeepHDRVideo数据集是由阿里巴巴达摩院创建,用于高动态范围(HDR)视频重建的基准数据集。该数据集包含97个静态场景序列和184个动态场景测试对,总计281个序列。数据集通过使用Basler acA4096-30uc相机捕捉,具有4096×2168的高分辨率和26fps的帧率。创建过程中,通过多曝光图像合并生成静态场景的HDR帧,动态场景则通过特殊方法生成HDR对。该数据集主要用于评估和改进HDR视频重建技术,特别是在处理不同曝光下的视频序列时的性能。
The DeepHDRVideo dataset is a benchmark dataset created by Alibaba DAMO Academy for high dynamic range (HDR) video reconstruction. This dataset contains 97 static scene sequences and 184 dynamic scene test pairs, totaling 281 sequences in all. The dataset was captured using a Basler acA4096-30uc camera, with a high resolution of 4096×2168 and a frame rate of 26 fps. During its creation, HDR frames for static scenes were generated via multi-exposure image fusion, while HDR pairs for dynamic scenes were produced using specialized methods. This dataset is primarily used to evaluate and improve HDR video reconstruction technologies, particularly the performance when handling video sequences under varying exposures.




