OBJECTFOLDER

OBJECTFOLDER是由斯坦福大学创建的一个包含100个虚拟对象的数据集，专注于多感官对象识别任务。该数据集通过统一的、以对象为中心的隐式表示，为每个对象编码视觉、听觉和触觉感官数据，使其在机器学习和机器人任务中具有广泛的应用潜力。数据集中的对象来自在线资源库，每个对象都通过视觉网络、音频网络和触觉网络进行模拟，以提供完整的感官概况。OBJECTFOLDER旨在通过实例识别、跨感官检索、3D重建和机器人抓取等基准任务，展示其在多感官感知和控制方面的实用性。

OBJECTFOLDER is a dataset consisting of 100 virtual objects created by Stanford University, focusing on multi-sensory object recognition tasks. It encodes visual, auditory, and tactile sensory data for each object through a unified, object-centric implicit representation, endowing it with extensive application potential in machine learning and robotics-related tasks. The objects in the dataset are sourced from online repositories, and each object is simulated via visual, audio, and tactile networks to provide a comprehensive sensory profile. OBJECTFOLDER aims to demonstrate its practical utility in multi-sensory perception and control through benchmark tasks such as instance recognition, cross-modal retrieval, 3D reconstruction, and robotic grasping.