SpatialSense
收藏资源简介:
SpatialSense是一个专注于空间关系识别的数据集,由普林斯顿大学创建。该数据集包含17,498个空间关系,分布在11,569张图像上,旨在为计算机视觉技术提供一个全面的测试平台。数据集通过对抗性众包方式构建,这种方法显著减少了数据集的偏差,并能从长尾中采样更多有趣的关系。SpatialSense的应用领域主要集中在提升计算机视觉系统的空间推理能力,解决复杂的空间关系识别问题。
SpatialSense is a dataset dedicated to spatial relation recognition, developed by Princeton University. It encompasses 17,498 spatial relations distributed over 11,569 images, and is designed to serve as a comprehensive testbed for computer vision technologies. The dataset is built through adversarial crowdsourcing, a method that notably reduces dataset bias and enables the sampling of more intriguing relations from the long tail. The primary application domains of SpatialSense focus on enhancing the spatial reasoning capabilities of computer vision systems and tackling complex spatial relation recognition tasks.




