HRSD

HRSD数据集是由德国人工智能研究中心创建的高分辨率合成深度数据集，包含100,000对彩色图像及其对应的密集地面真值深度图，分辨率为1920×1080。该数据集通过使用商业视频游戏《侠盗猎车手V》（GTA-V）生成，涵盖了从室内到室外，从均匀表面到纹理的多样化场景。数据集的创建旨在为单目深度估计（MDE）算法提供高质量的训练数据，以解决现有数据集在分辨率和一致性方面的不足。HRSD数据集的应用领域包括自动驾驶、场景重建和点云创建等，旨在提高MDE算法在真实世界场景中的准确性和泛化能力。

The HRSD (High-Resolution Synthetic Depth) dataset was created by the German Research Center for Artificial Intelligence. It encompasses 100,000 pairs of color images and their corresponding dense ground-truth depth maps, with a resolution of 1920×1080. Generated using the commercial video game Grand Theft Auto V (GTA-V), this dataset covers diverse scenarios ranging from indoor to outdoor environments and from uniform surfaces to textured ones. The dataset was developed to provide high-quality training data for monocular depth estimation (MDE) algorithms, addressing the limitations of existing datasets in terms of resolution and consistency. Its application fields include autonomous driving, scene reconstruction, point cloud generation and other related areas, aiming to improve the accuracy and generalization capability of MDE algorithms in real-world scenarios.