Austin Mutex
收藏资源简介:
MUTEX 数据集由美国德克萨斯大学奥斯汀分校的研究团队创建,旨在支持多模态机器人任务规范的研究。该数据集包含 100 个模拟任务和 50 个真实世界任务,涵盖多种操作场景,如“将面包放在烤盘上”和“打开空气炸锅放入热狗”。每个任务均配备 11 种不同模态的任务规范,包括视频演示、目标图像、文本指令、文本目标描述、语音指令和语音目标描述,总计超过 6000 条标注数据。数据集的创建过程结合了模拟环境和真实世界场景,通过人类演示和自然语言生成技术生成多样化任务规范。其应用领域主要集中在机器人学习和人机交互,旨在提升机器人对多模态任务指令的理解与执行能力,推动机器人在复杂环境中的自主操作研究。
The MUTEX dataset was developed by a research team from The University of Texas at Austin, with the core objective of supporting research on multimodal robotic task specification. This dataset comprises 100 simulated tasks and 50 real-world tasks, covering diverse manipulation scenarios including "placing bread on a baking tray" and "opening an air fryer to insert a hot dog". Each task is paired with 11 distinct modalities of task specifications, namely video demonstrations, target images, text instructions, text target descriptions, speech instructions and speech target descriptions, resulting in a total of over 6,000 annotated data samples. The construction of this dataset integrates simulated environments and real-world scenarios, generating varied task specifications via human demonstrations and natural language generation technologies. Its primary application domains lie in robotic learning and human-robot interaction, aiming to enhance robots' ability to comprehend and execute multimodal task instructions, and advance research on autonomous robotic operation in complex environments.




