• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    AI

    Chris Paxton Shares VR Robotic Glove Data Tool Video

    Paxton shared a split-screen video of VR-controlled robotic gloves handling lab equipment.

    CP
    RP
    ML
    5 Sources, 28d ago, first seen 28d ago

    TLDR

    Chris Paxton, an AI and robotics researcher who leads AI at Agility Robotics, posted on X praising a video for high quality scalable dexterous data collection. The video from XSquareRobot shows split-screen footage of a man in a black t-shirt wearing a VR headset and white three-fingered robotic gloves while manipulating lab equipment. Paxton described the setup as a great tool. The post includes a link to the 43-second video.

    Combined views

    20.5K

    5 Sources, first seen 28d ago

    Combined views

    20.5K

    5 Sources, first seen 28d ago

    117 likes
    117 likes
    9 comments
    54 saves
    81 reposts

    Sentiment

    Positive98%2%Negative

    Based on 51 sentiment-bearing replies from 51 accounts across 2 conversations.

    9 comments
    54 saves
    81 reposts

    Sentiment

    Positive98%2%Negative

    Based on 51 sentiment-bearing replies from 51 accounts across 2 conversations.

    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet

    5 Sources

    @chris_j_paxtonSeeing that completing long horizon tasks is a data problem -- that with good tools like these co-designed isomorphic dexterous data collection tools, you can accomplish long horizon tasks with less data than we might have expected
    @rohanpaul_aiDexterous robots have a data problem that gets expensive very quickly, and recently, so many brilliant data projects are coming to resolve this problem. X Square Robot released TwinDEX, a matched wearable-and-robotic hand system that lets humans collect robot training data without using a robot. This will potentially remove one of the biggest limits on scaling dexterous robot learning. Claims up to 5.3X the effective collection throughput of on-robot teleoperation TwinDEX combines a wearable three-finger data-collection hand with a closely matched robotic hand so human demonstrations can transfer directly into robot control without collecting them on the robot itself.
    @liu_mingyuRT @XSquareRobot: What if a dexterous robot could learn a 20+ step chemistry experiment without a single on-robot training demo? Meet Twin…

    5 Sources

    @chris_j_paxtonSeeing that completing long horizon tasks is a data problem -- that with good tools like these co-designed isomorphic dexterous data collection tools, you can accomplish long horizon tasks with less data than we might have expected
    @rohanpaul_aiDexterous robots have a data problem that gets expensive very quickly, and recently, so many brilliant data projects are coming to resolve this problem. X Square Robot released TwinDEX, a matched wearable-and-robotic hand system that lets humans collect robot training data without using a robot. This will potentially remove one of the biggest limits on scaling dexterous robot learning. Claims up to 5.3X the effective collection throughput of on-robot teleoperation TwinDEX combines a wearable three-finger data-collection hand with a closely matched robotic hand so human demonstrations can transfer directly into robot control without collecting them on the robot itself.
    @liu_mingyuRT @XSquareRobot: What if a dexterous robot could learn a 20+ step chemistry experiment without a single on-robot training demo? Meet Twin…