SANPO
收藏资源简介:
SANPO数据集是由谷歌研究团队创建的大型户外场景理解数据集,专注于密集预测任务。该数据集包含112,000个真实视频和113,000个合成视频的密集分割掩码,以及相应的深度图。数据集采集自美国多个地点,覆盖了不同的环境、天气条件和时间。SANPO数据集不仅支持人类导航任务,还适用于多种密集预测任务,旨在推动视频分割、深度估计和合成到真实域适应等领域的技术发展。
The SANPO dataset is a large-scale outdoor scene understanding dataset created by the Google Research team, focusing on dense prediction tasks. It contains dense segmentation masks and corresponding depth maps for 112,000 real-world videos and 113,000 synthetic videos. The dataset is collected from multiple locations across the United States, covering diverse environments, weather conditions and time periods. Beyond supporting human navigation tasks, the SANPO dataset is applicable to a variety of dense prediction tasks, aiming to advance technological developments in fields such as video segmentation, depth estimation and synthetic-to-real domain adaptation.

- 1SANPO: A Scene Understanding, Accessibility, Navigation, Pathfinding, Obstacle Avoidance Dataset谷歌研究 · 2023年



