非中心全景室内数据集
收藏资源简介:
非中心全景室内数据集是由阿拉贡工程研究学院计算机科学与系统工程系创建的,包含2574张RGB非中心全景图,涵盖约650个不同室内环境。数据集不仅包含全景图像,还附带深度图和用于房间布局获取的标注信息。该数据集通过随机生成虚拟环境,利用特定的相机投影模型进行图像渲染,实现像素级自动标注。主要应用于室内场景理解,解决基于学习的场景理解算法中数据不足的问题,特别适用于利用非中心系统几何特性的算法开发和评估。
The Non-Central Panoramic Indoor Dataset was developed by the Department of Computer Science and Systems Engineering, Aragon Institute of Engineering Research. It contains 2574 RGB non-central panoramic images spanning approximately 650 unique indoor environments. In addition to the panoramic images, the dataset also includes depth maps and annotation information for room layout extraction. This dataset achieves pixel-level automatic annotation by randomly generating virtual environments and rendering images using a dedicated camera projection model. Its primary application scenario is indoor scene understanding, addressing the issue of insufficient training data for learning-based scene understanding algorithms, and it is particularly suitable for the development and evaluation of algorithms that exploit the geometric properties of non-central panoramic systems.

- 1Non-central panorama indoor dataset阿拉贡工程研究学院计算机科学与系统工程系 · 2024年



