EF-SAI
收藏资源简介:
我们完全记录了988组的配对数据集,其中包含30个APS帧和收集有遮挡的并发事件,以及1个无遮挡的APS帧作为地面真相。我们提出的EF-SAI数据集在大小和品种方面都有所不同,其中遮挡密度从稀疏到极密集不等。因此,可以将ef-sai数据集用于在不同密度的遮挡下训练和评估基于帧和基于事件的SAI算法。我们的EF-SAI数据集可在https://github.com/smjsc/EF-SAI获得。
We have fully compiled and documented 988 sets of paired data, where each set includes 30 APS frames paired with occluded concurrent events, alongside one unoccluded APS frame as the ground truth. The proposed EF-SAI dataset exhibits variability in both scale and diversity, with occlusion densities ranging from sparse to extremely dense. Accordingly, the EF-SAI dataset can be utilized to train and evaluate both frame-based and event-based SAI algorithms under varying occlusion densities. Our EF-SAI dataset is publicly available at https://github.com/smjsc/EF-SAI.




