InteriorNet
收藏资源简介:
InteriorNet是一个大规模、多传感器、照片级真实感的室内场景数据集,由帝国理工学院和KooLab联合创建。该数据集包含约100万个家具CAD模型和2200万个室内布局,这些数据来源于真实的生产和装饰。数据集分为两部分:15000个序列,每个序列包含1000张图像;500万张图像,每个布局包含3张图像。数据集通过模拟真实场景变化,如物体移动和光照变化,来增强其真实感。InteriorNet主要用于SLAM(同时定位与地图构建)算法的训练和评估,以及室内场景理解的基准测试。
InteriorNet is a large-scale, multi-sensor, photorealistic indoor scene dataset jointly created by Imperial College London and KooLab. This dataset contains approximately 1 million furniture CAD models and 22 million indoor layouts sourced from real-world production and decoration scenarios. The dataset is divided into two parts: 15,000 sequences each containing 1,000 images, and 5 million images with 3 images per layout. To enhance realism, the dataset simulates real-world scene variations such as object movements and lighting changes. InteriorNet is primarily used for the training and evaluation of SLAM (Simultaneous Localization and Mapping) algorithms, as well as benchmark testing for indoor scene understanding.




