Shanghai Debuts Open-Source Dataset to Accelerate Humanoid Robot Evolution

Shanghai launched the first and largest cross-embodiment, vision-based multi-modal dataset for embodied intelligence and humanoid robotics.

The National and Local Co-Build Humanoid Robotics Innovation Center and Shanghai startup ViTai Robotics created the "White Tiger" VTouch Dataset. At least 6,000 of the 60,000 minutes are on OpenLoong open-source community.

VTouch uses joint pose data and vision-based tactile sensors, unlike other datasets. It has humanoid and wheeled robots. This event follows the center's previous launch of the Qingloong ("Azure Dragon") humanoid robot. Both projects, named after Chinese mythical beasts, mark a new era for domestic robotics.

The launch comes as Shanghai integrates humanoid robotics into its long-term economic blueprint. Shanghai's economy includes humanoid robotics with the launch. At the Shanghai Two Sessions, Shanghai Municipal People's Congress deputy and Guochuang Zhizao Technology executive director He Renlong stressed the industry's critical moment.

Shanghai humanoid robotics is moving from "technology validation" to "scalable application," He said. He warned that core technological breakthroughs, industrial collaboration, and standardized application ecosystems are hard.

For years, the industry has struggled to find high-quality, real-world physical interaction data, agile robot training "fuel." Shanghai hopes to solve this bottleneck by open-sourcing VTouch to give researchers diverse interaction data for humanoid robots.