Xingyuan Releases Robot Data Engine Chorus: Single Human Demonstration Can Be Reused Across Embodiments, Achieving Over 40x Data Expansion
行元发布机器人数据引擎Chorus：单份人类示范可跨本体复用，实现超 40 倍数据扩容
Xingyuan Intelligent has released the cross-embodiment data engine Chorus, which supports converting human operation videos collected from various devices into pre-trained data suitable for different robot embodiments, achieving over 40 times data expansion. The engine improves data reuse efficiency and reduces training costs through a unified data model and automated processing pipeline.
The release highlights a cross-embodiment data engine designed to convert human demonstration videos into robot training data, with claims of 40x data expansion and 85% conversion efficiency. The tool is positioned as a key component in the company's System Scaling approach to embodied intelligence.
Source: Chinese robotics — Research news discovery · Read original article ↗
English translation unavailable. Choose the original article to read it.
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
data expansion
40
cost per hour of trajectory processing
15
View original evidence
在当前测试配置下，Chorus 通过 16 种机器人本体适配，将每小时有效机器人轨迹的处理成本降至 15 元；结合 3 倍指令增广，可实现约 40 倍的有效训练数据扩容，单位训练样本的等效成本仅需 5 元/小时。
Open source S12cost per hour of training sample
5
View original evidence
在当前测试配置下，Chorus 通过 16 种机器人本体适配，将每小时有效机器人轨迹的处理成本降至 15 元；结合 3 倍指令增广，可实现约 40 倍的有效训练数据扩容，单位训练样本的等效成本仅需 5 元/小时。
Open source S12data expansion with instruction augmentation
≈40
View original evidence
在当前测试配置下，Chorus 通过 16 种机器人本体适配，将每小时有效机器人轨迹的处理成本降至 15 元；结合 3 倍指令增广，可实现约 40 倍的有效训练数据扩容，单位训练样本的等效成本仅需 5 元/小时。
Open source S12
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
行元发布机器人数据引擎Chorus：单份人类示范可跨本体复用，实现超 40 倍数据扩容
Open source S1
数据处理过程中，Chorus通过三维运动重建、轨迹优化与质量校验，减少相机移动、手部遮挡等因素对转换结果的影响，将平均可用的数据转化率提升至85%。
Open source S11
在当前测试配置下，Chorus 通过 16 种机器人本体适配，将每小时有效机器人轨迹的处理成本降至 15 元；结合 3 倍指令增广，可实现约 40 倍的有效训练数据扩容，单位训练样本的等效成本仅需 5 元/小时。
Open source S12
其次，与业内“一份数据对应一种机器人”的普遍方案不同，Chorus可将单份人类示范转化为适配不同机器人身体结构和运动范围的训练数据，实现跨本体复用。
Open source S9
首先，Chorus支持智能眼镜、胸戴相机、腕部相机、UMI类设备等多种设备采集的人类操作视频，将不同硬件、视角和采集条件下符合处理要求的数据纳入统一流程，降低对单一采集设备的依赖，扩大可用数据来源。
Open source S8
Source：Chinese robotics — Research news discovery · jiemian.com