ModelNet40‡, McGill‡

ModelNet40‡和McGill‡是专为开放姿态3D零样本分类设计的基准数据集。这些数据集通过对ModelNet40和McGill数据集中的每个样本应用随机旋转来创建，旨在模拟现实世界中物体可能出现的任意姿态。这些数据集不仅包含多种常见物体类别，如椅子、桌子、飞机等，还包括生物样本如鱼、恐龙、蜘蛛等，适用于评估和推动3D零样本分类技术的发展，特别是在处理非标准姿态的3D对象识别方面的挑战。

ModelNet40‡ and McGill‡ are benchmark datasets specifically designed for open-pose 3D zero-shot classification. These datasets are generated by applying random rotations to each sample from the original ModelNet40 and McGill datasets, aiming to simulate arbitrary poses that real-world objects may take. They cover a wide range of common object categories such as chairs, tables, airplanes, and the like, as well as biological specimens including fish, dinosaurs, spiders, and others. These datasets are suitable for evaluating and advancing the development of 3D zero-shot classification technologies, particularly addressing the challenges in 3D object recognition under non-standard poses.