The article introduces SANPO, a large-scale egocentric video dataset aimed at aiding research in outdoor human egocentric scene understanding. The dataset includes both real and synthetic data, with panoptic segmentation and depth annotations. It is designed to assist in tasks such as video segmentation, depth estimation, visual modeling, and synthetic-to-real domain adaptation. The authors believe that SANPO will help advance the state-of-the-art in these areas and enable better human navigation systems.
Publication date: 21 Sep 2023
Project Page: https://google-research-datasets.github.io/sanpo_dataset/
Paper: https://arxiv.org/pdf/2309.12172
