OmniSharing DB 帕西尼全模态具身智能数据集

OmniSharing DB 帕西尼全模态具身智能数据集是国内首个融合触觉、视觉、语言与动作四维模态的高精度机器人训练数据资产，依托帕西尼感知科技总部独创的PMEC超采技术与VTLA（视觉-触觉-语言-动作）模型框架构建。该资产通过向总部采购自研的触觉手套，以及多维视觉阵列及语音系统等数据采集设备，完整记录人类在工业装配、家居服务、特种作业等各类场景下的物理交互过程，覆盖触觉力反馈（分布力/合力大小与方向）、关节角度轨迹、多光谱视觉流（RGB/深度/红外图像）、语音指令及文本任务描述，并附加场景布局、动作阶段标签等结构化标注。目前单日采集量已突破50万条，计划年采集近2亿条数据。采集的数据经匿名化与标准化处理，具备毫米级力学保真度与跨本体泛化能力，为具身智能体提供“人类级”动作模仿与物理直觉训练基座。该数据资产直接响应了2025年政府工作报告关于“培育具身智能新兴产业”的政策方向，极具经济价值与行业代表性。

OmniSharing DB: Pasini Multimodal Embodied Intelligence Dataset is China's first high-precision robotic training data asset integrating four modalities: tactile, visual, linguistic and motor. It is constructed based on the independently developed PMEC ultra-collection technology and VTLA (Vision-Tactile-Language-Action) model framework from the headquarters of Pasini Perception Technology. This asset fully records the physical interaction processes of humans in various scenarios such as industrial assembly, home services and special operations, by purchasing self-developed tactile gloves, multi-dimensional visual arrays, voice systems and other data collection equipment from the headquarters. It covers tactile force feedback (distributed force/resultant force magnitude and direction), joint angle trajectories, multispectral visual streams (RGB/depth/infrared images), voice commands and text task descriptions, with additional structured annotations including scene layout and action stage labels. Currently, the daily collection volume has exceeded 500,000 entries, with an annual collection target of nearly 200 million data entries. The collected data has been anonymized and standardized, featuring millimeter-scale mechanical fidelity and cross-embodiment generalization capability, providing a foundational training platform for embodied AI agents to conduct "human-level" action imitation and physical intuition training. This data asset directly aligns with the policy direction of "fostering the emerging embodied intelligence industry" put forward in the 2025 Government Work Report, and possesses considerable economic value and industry representativeness.