ScanObjectNN
收藏资源简介:
ScanObjectNN是由香港科技大学等机构创建的一个新的真实世界点云对象数据集,基于扫描的室内场景数据。该数据集包含2902个对象,分为15个常见类别,数据来源于SceneNN和ScanNet两个场景网格数据集。创建过程中,研究者手动筛选和选择对象,并考虑额外的对象扰动以丰富数据集。ScanObjectNN旨在解决真实世界对象分类的挑战,如背景杂乱和部分遮挡,适用于3D计算机视觉中的对象分类和分割任务。
ScanObjectNN is a novel real-world point cloud object dataset developed by institutions including the Hong Kong University of Science and Technology, based on scanned indoor scene data. This dataset contains 2902 objects, categorized into 15 common classes, and is sourced from two scene mesh datasets, SceneNN and ScanNet. During its creation, researchers manually screened and selected objects, and introduced additional object perturbations to enrich the dataset. ScanObjectNN aims to address the challenges of real-world object classification, such as background clutter and partial occlusion, and is applicable to object classification and segmentation tasks in 3D computer vision.

- 1Revisiting Point Cloud Classification: A New Benchmark Dataset and Classification Model on Real-World Data香港科技大学 · 2019年



