计算机视觉数据集|室内导航数据集

NYU-Depth V2数据集包含1449张配准的RGB-D图像，来自464个不同室内场景，每张图像均带有深度信息和密集的多类别标注，涵盖40余类物体实例。数据集规模适中，标注精细，包含预处理后的完整深度图与语义标签，适用于场景理解、深度估计、语义分割等任务。其广泛应用于计算机视觉与深度学习研究，尤其适合室内环境感知相关模型的训练与评估。

The NYU-Depth V2 dataset comprises 1449 registered RGB-D images sourced from 464 unique indoor scenes. Each image is paired with depth information and dense multi-class annotations, covering more than 40 object instances. With a moderate scale and fine-grained annotations, the dataset provides complete preprocessed depth maps and semantic labels, which is applicable to tasks such as scene understanding, depth estimation and semantic segmentation. It has been widely used in computer vision and deep learning research, and is particularly suitable for training and evaluating models related to indoor environment perception.