OBJECTFOLDER 2.0

OBJECTFOLDER 2.0是由斯坦福大学和卡内基梅隆大学合作创建的一个大规模多感官对象数据集，旨在促进计算机视觉和机器人学中的多感官学习。该数据集包含1000个常见家用物品的隐式神经表示，涵盖视觉、听觉和触觉数据，显著提升了数据集规模和渲染速度。通过高质量的多感官渲染，该数据集支持模型从虚拟对象到真实世界的成功转移，特别适用于对象尺度估计、接触定位和形状重建等任务。OBJECTFOLDER 2.0为多感官学习提供了一个新的路径和测试平台，推动了计算机视觉和机器人学领域的技术发展。

OBJECTFOLDER 2.0 is a large-scale multi-sensory object dataset co-created by Stanford University and Carnegie Mellon University, aiming to advance multi-sensory learning research in computer vision and robotics. This dataset includes implicit neural representations of 1000 common household items, covering visual, auditory and tactile data, and significantly improves both dataset scale and rendering speed. Through high-quality multi-sensory rendering, it supports the successful transfer of models from virtual objects to the real world, and is particularly applicable to tasks such as object scale estimation, contact localization and shape reconstruction. OBJECTFOLDER 2.0 provides a novel pathway and testbed for multi-sensory learning, driving technological progress in the fields of computer vision and robotics.

搜集汇总

背景与挑战

背景概述

ObjectFolder 2.0是一个包含1000个对象的多感官数据集，每个对象通过隐式神经表示网络（VisionNet、AudioNet、TouchNet）支持视觉、听觉和触觉数据的渲染。数据集适用于模拟到现实的转换研究，遵循CC-BY-4.0许可证，并提供详细的下载和渲染指南。