VOID

该数据集名为VOID，包含了室内外场景的同步RGB图像和稀疏深度图。这些稀疏深度图大约含有1500个稀疏深度点，密度约为0.5%。此外，通过主动立体技术获得了密集的真实深度图，深度值范围在0.2米至5.0米之间。数据集规模包括56个序列，其中45000帧用于训练，以及8个序列，800帧用于测试。该数据集的任务是深度补全。

This dataset, named VOID, contains synchronized RGB images and sparse depth maps from both indoor and outdoor scenes. Each of these sparse depth maps contains approximately 1,500 sparse depth points, with a density of around 0.5%. In addition, dense ground-truth depth maps obtained via active stereo technology are provided, with depth values ranging from 0.2 meters to 5.0 meters. The dataset includes 56 sequences, among which 45,000 frames are used for training, and another 8 sequences totaling 800 frames are reserved for testing. The core task supported by this dataset is depth completion.