ROVR-Open-Dataset
收藏资源简介:
ROVR-Open-Dataset是一个大规模的深度数据集,专为自动驾驶领域设计。该数据集包含了20K个视频帧,用于评估现有的深度估计方法。数据集的采集过程采用了轻量级的采集流程,确保在低成本的情况下广泛覆盖场景。此外,该数据集的深度真值虽然稀疏但统计上足够,使得深度估计器能够进行稳健的训练。与现有的数据集相比,ROVR-Open-Dataset在驾驶场景的多样性和深度密度方面更具优势,为深度估计研究提供了新的挑战。
ROVR-Open-Dataset is a large-scale depth dataset specifically designed for the autonomous driving domain. It contains 20,000 video frames for evaluating existing depth estimation methods. The dataset adopts a lightweight acquisition workflow, ensuring extensive coverage of various scenarios at low cost. Furthermore, although its ground-truth depth data is sparse, it is statistically sufficient to enable robust training of depth estimators. Compared with existing datasets, ROVR-Open-Dataset has advantages in both driving scene diversity and depth density, thus providing new challenges for depth estimation research.

- 1通过ROVR Labs, Inc. · 2025年



