Baihu-VTouch Visuo-Tactile Dataset

A cross-embodiment visuotactile dataset released in 2026 by China's National Humanoid Robot Innovation Center and Weitai Robotics.

Baihu-VTouch was released on January 26, 2026 by the National and Local Co-built Humanoid Robotics Innovation Center (Shanghai) together with Shanghai Weitai Robotics, a visuotactile-sensor company. A visuotactile sensor uses a built-in camera to photograph the deformation of a soft gel surface as it's pressed, sensing the shape and force of contact. According to the release, the dataset totals over 60,000 minutes and about 90.72 million real object-contact samples, recording visuotactile images, RGB-D footage, and joint poses in sync; the embodiments used include wheeled-arm robots, the bipedal humanoid Qinglong, and handheld smart terminals. Tasks span 4 scene categories — household chores, industrial manufacturing, food service, and specialized operations — with more than 380 tasks, over 100 atomic skills, and more than 500 real objects in total. Most existing large-scale robot data has only vision and action; this dataset adds the missing dimension of contact. The first 6,000 minutes have already been released on the OpenLoong open-source community.

ExampleThe dataset includes more than 260 contact-intensive tasks; when training a manipulation policy that needs a sense of “feel,” these visuotactile images can be used as input alongside vision and pose.

Also called

VTouch

Related

Vision-Based Tactile Sensor · Tactile Data · Cross-Embodiment Data · Weitai Robotics · Qinglong (OpenLoong) · National and Local Co-built Humanoid Robotics Innovation Center (Shanghai)

Sources

国地中心发布全球首个跨本体视触觉多模态数据集「白虎-VTouch」（腾讯新闻） (Chinese)

As of

2026-01