Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Video

Updated: May 9, 2026

Capturing Dynamic Finger Gesturing with High-resolution Surface Electromyography and Computer Vision
08:15

Capturing Dynamic Finger Gesturing with High-resolution Surface Electromyography and Computer Vision

Published on: March 28, 2025

DynamicTHOR: A Scalable Dataset of Human-Centric Dynamic Scenes for Embodied AI.

Chenxu Wang1, Dunzheng Wang1, Dong Wang1

  • 1Department of Computer Science and Technology, BNRIST, Institute for Embodied Intelligence and Robotics, Tsinghua University, Beijing, China.

Scientific Data
|May 7, 2026
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

Virtual Work for a System of Connected Rigid Bodies01:06

Virtual Work for a System of Connected Rigid Bodies

Virtual work is a powerful method used to solve problems involving several connected rigid bodies. When the system is in equilibrium, virtual work is zero. This allows the calculation of the resulting forces when a system undergoes a virtual displacement. When attempting to analyze such a system, first, use a free-body diagram, where an independent coordinate represents the configuration of the links, and mark its deflected position resulting from the positive virtual displacement.
Next,...

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

An Electrical Capacitance Tomography Dataset for Image Reconstruction Benchmarking.

Scientific data·2026
Same author

Proactive collaboration via autonomous interaction.

Nature communications·2026
Same author

S2R-Bench: A Sim-to-Real Evaluation Benchmark for Autonomous Driving.

Scientific data·2025
Same author

A Visual Dataset for Anomaly Detection in Self-Driving Laboratories.

Scientific data·2025
Same author

A simulated dataset for proactive robot task inference from streaming natural language dialogues.

Scientific data·2025
Same author

Pyrolytic hydrocarbon growth from cyclopentadiene.

The journal of physical chemistry. A·2010
Same journal

High-resolution thermal infrared dataset for airborne person detection in SAR missions.

Scientific data·2026
Same journal

USV-derived bathymetry of high-risk glacial lakes and a critical semi-arid ecosystem lake in the Himalaya.

Scientific data·2026
Same journal

A large-scale, LLM-assisted and validated dataset of biomass and waste conversion technologies and feedstocks.

Scientific data·2026
Same journal

Near-complete telomere-to-telomere genome assembly of a male barbel steed (Hemibarbus labeo).

Scientific data·2026
Same journal

A near telomere-to-telomere genome assembly of Rhodiola macrocarpa (Crassulaceae).

Scientific data·2026
Same journal

SowPostureDS: A Multi-Class Image Dataset for YOLO-Based Detection of Sow Postures in diverse Farrowing Systems.

Scientific data·2026
See all related articles

This study introduces a framework using Large Language Models (LLMs) to simulate human activities and generate dynamic scenes for embodied AI research. The resulting DynamicTHOR dataset offers comparable quality to human annotations, overcoming previous data limitations.

Area of Science:

  • Robotics
  • Artificial Intelligence
  • Computer Vision

Background:

  • Existing research on dynamic environments is limited by dataset scale due to high annotation costs.
  • Simulating human activities and object dynamics in daily life scenarios is challenging.

Purpose of the Study:

  • To introduce a novel framework for generating dynamic embodied scenes using Large Language Models (LLMs).
  • To create the DynamicTHOR dataset, featuring simulated human activities and object dynamics.
  • To overcome limitations in dataset scale and annotation costs for dynamic environment research.

Main Methods:

  • Utilized Large Language Models (LLMs) to simulate human activities and object dynamics.
  • Applied simulated human residents to embodied scenes to generate dynamic environments.

Related Experiment Videos

Last Updated: May 9, 2026

Capturing Dynamic Finger Gesturing with High-resolution Surface Electromyography and Computer Vision
08:15

Capturing Dynamic Finger Gesturing with High-resolution Surface Electromyography and Computer Vision

Published on: March 28, 2025

  • Crafted the DynamicTHOR dataset with 50 characters and 100 dynamic scenes.

    • Main Results:

    • Successfully generated dynamic scenes with believable and diversified object movements.
    • The DynamicTHOR dataset demonstrated quality comparable to human annotations.
    • Validated the framework's effectiveness through a user study comparing generated dynamics against baseline and human data.

    Conclusions:

    • The LLM-based framework effectively generates high-quality dynamic scenes for embodied AI.
    • The DynamicTHOR dataset provides a scalable solution for studying AI in dynamic environments.
    • This work facilitates advancements in embodied intelligence, particularly for navigation tasks in complex, changing scenarios.