Depth360
收藏资源简介:
Depth360数据集是由早稻田大学科学与工程研究所创建的大规模数据集,旨在解决单视图深度估计训练数据的稀缺问题。该数据集包含30,000条数据,通过利用互联网上的360度视频,采用测试时训练方法,结合独特的几何和时间约束,生成一致且令人信服的深度样本。Depth360数据集特别适用于单视图深度估计,尤其是在自主驾驶和场景重建等应用领域,旨在提高对环境理解的准确性。
The Depth360 dataset is a large-scale dataset developed by the Research Institute for Science and Engineering of Waseda University, which is designed to tackle the shortage of training data for single-view depth estimation tasks. It contains 30,000 data samples, which are generated by utilizing 360-degree videos sourced from the Internet, adopting test-time training methodologies, and integrating unique geometric and temporal constraints to produce consistent and credible depth samples. The Depth360 dataset is particularly well-suited for single-view depth estimation, especially in application domains such as autonomous driving and scene reconstruction, with the aim of enhancing the accuracy of environmental understanding.

- 1360 Depth Estimation in the Wild -- The Depth360 Dataset and the SegFuse Network早稻田大学科学与工程研究所 · 2022年



