UrbanLF: A Comprehensive Light Field Dataset for Semantic Segmentation of Urban Scenes

As one of the fundamental technologies for scene understanding, semantic segmentation has been widely explored in last few years. Light field cameras encode geometric information by simultaneously recording spatial and angular light rays, which provides us with a new way to solve this issue. In paper, we propose high-quality challenging urban dataset, containing 1074 samples composed real-world synthetic images as well pixel-wise annotations 14 classes. To best our knowledge, it is largest most diverse dataset segmentation. We further design two baselines tailored compare them state-of-the-art RGB, video RGB-D-based methods using proposed dataset. The outperforming results demonstrate advantages task. also provide evaluations super-resolution depth estimation methods, showing that presents challenges supports detailed comparisons among different methods. expect work inspires research direction stimulates scientific progress related fields. complete available at https://github.com/HAWKEYE-Group/UrbanLF .