广告
加载中

星忆智能完成数千万元种子轮及种子+轮融资 以高质量数据构建具身智能物理AI基座

龚作仁 2026-10-22 15:38
龚作仁 2026/10/22 15:38

邦小白快读

EN
全文速览

总1 融资与公司定位

1. 星忆智能完成数千万元种子轮及种子+轮连续融资,种子轮由顺禧基金领投,种子+轮由凌云光领投,中关村资本、一创投资跟投。

2. 资金重点投入核心算法与产品研发、真实世界多模态数据基础设施建设、客户交付与商业化能力提升。

总2 数据业务干货

1. 具身智能进入任务落地阶段,复杂手部操作原始Ego数据难以直接转成训练数据,高质量、可学习、可规模化交付的训练数据才是瓶颈。

2. 公司以EgoKit多模态采集系统与HBR Engine智能数据引擎,针对严重遮挡、图像畸变、高速运动、极端光照,做手部姿态解析、多模态融合处理与自动化标注,把真实人类操作加工为可复用训练资产。

总3 合作与团队

1. 已与多家头部具身模型企业及机器人厂商开展项目级合作,输出高精度手部交互训练数据,沉淀标准化数据交付能力。

2. 公司孵化于清华计算机系,核心团队来自清华、北航,兼具机器人、AI、自动驾驶产业经验,发表CVPR、NeurIPS等顶会论文80余篇。

总1 品牌定位与背书

1. 星忆智能定位为具身智能物理AI数据基础设施服务商,孵化于清华计算机系,以算法为核心壁垒、数据质量为价值落点,形成技术型品牌标签。

2. 种子轮由顺禧基金领投,种子+轮由凌云光领投,中关村资本、一创投资跟投;市场化VC、产业龙头与国资引导基金联合入局,为技术稀缺性提供背书。

总2 产品研发与差异化

1. 自研EgoKit多模态采集系统与HBR Engine智能数据引擎,针对严重遮挡、图像畸变、高速运动、极端光照四大痛点,构建手部姿态解析、多模态融合处理与自动化标注能力。

2. 面向复杂手物交互,把真实人类操作加工为高质量、可复用、可规模化交付训练资产,以稀缺数据和标准化交付构建B端差异化。

总3 渠道与产业协同

1. 凌云光既是产业投资方也是业务协同伙伴,通过全资子公司元客视界与星忆智能战略合作,整合第一视角Ego采集、多视点视觉采集、光学动捕硬件能力。

2. 已与多家头部具身模型企业及机器人厂商项目级合作,输出高精度手部交互训练数据,沉淀标准化数据交付能力,体现产业渠道与交付能力协同。

总1 市场机会与增长信号

1. 具身智能迈入任务落地阶段,复杂手物交互的高质量训练数据属于行业稀缺供给,手部训练数据存在巨大市场缺口。

2. 星忆聚焦第一人称真实交互数据,建设全栈数据基础设施,已与多家头部具身模型企业及机器人厂商项目级合作,显示B端需求与合作机会。

总2 可学习商业模式与合作方式

1. 以算法为核心壁垒、数据质量为价值落点,通过EgoKit采集、HBR Engine处理、手部姿态解析、多模态融合、自动化标注,形成采集、标注、质量评估、训练数据交付闭环。

2. 凌云光以CVC产业资本领投,并通过元客视界战略合作,整合硬件采集与自动化数据处理技术,形成技术能力+产业场景协同;国资耐心资本支持硬科技成果产业化转化。

总3 风险与应对提示

1. 原始Ego数据并不缺少,但复杂手部操作涉及精细动作、频繁遮挡与复杂手物交互,难以直接转化为有效训练数据,高质量、可规模化交付才是瓶颈。

2. 团队需提升工程落地与商业化能力,文章提示具身智能竞争本质是高质量物理交互数据供给能力比拼。

总1 产品设计与生产需求

1. 针对严重遮挡、图像畸变、高速运动、极端光照四大技术痛点,需要适配复杂成像与动态交互场景的算法体系。

2. 自研EgoKit多模态采集系统与HBR Engine智能数据引擎,融合视觉、触觉、惯性等多源信息,支持长序列、强接触及人手—物体交互场景下的毫米级三维手部姿态解析与精细化标注。

总2 商业机会

1. 具身智能手部训练数据存在市场缺口,星忆已与多家头部具身模型企业及机器人厂商项目级合作,输出高精度手部交互训练数据。

2. 凌云光通过元客视界与星忆战略合作,整合第一视角Ego采集、多视点视觉采集、光学动捕硬件能力,为高端数据采集矩阵和工程壁垒带来协同机会。

总3 数字化与电商启示

1. 把真实人类操作工业化加工为高质量、可复用、可规模化交付训练资产,依赖自动化标注、多模态融合处理、质量评估和标准化数据交付闭环。

2. 对制造类企业而言,可学习以数据采集设备、算法处理和标准化交付打通从场景数据到训练资产的链路。

总1 行业趋势

1. 具身智能进入任务落地阶段,行业对真实世界操作数据提出更高标准,高质量、可学习、可规模化交付训练数据成为模型能力上限瓶颈。

2. 物理AI数据基础设施服务商出现,星忆以算法为核心壁垒、数据质量为价值落点,聚焦第一人称真实交互数据。

总2 新技术与解决方案

1. EgoKit多模态采集系统与HBR Engine智能数据引擎,针对严重遮挡、图像畸变、高速运动、极端光照,提供手部姿态解析、多模态融合处理与自动化标注能力。

2. 将真实人类操作加工为高质量、可复用、可规模化交付训练资产,形成采集、标注、质量评估、训练数据交付完整业务闭环。

总3 客户痛点与服务机会

1. 客户痛点是原始Ego数据多,但复杂手部精细动作、频繁遮挡、复杂手物交互数据难直接训练,稀缺的是可规模化交付的高质量数据。

2. 星忆已与多家头部具身模型企业及机器人厂商项目级合作,输出高精度手部交互训练数据,沉淀标准化数据交付能力;与元客视界硬件协同可增强服务能力。

总1 平台化需求与问题

1. 具身智能任务落地要求真实世界操作数据,但原始Ego数据难以直接转化为训练数据,高质量、可学习、可规模化交付的训练数据是瓶颈。

2. 平台需要解决复杂手部操作中的精细动作、频繁遮挡、复杂手物交互等数据加工问题,并建立质量评估与标准化交付能力。

总2 最新做法与运营管理

1. 星忆以算法为核心壁垒,自研EgoKit多模态采集系统与HBR Engine智能数据引擎,针对四大技术痛点做手部姿态解析、多模态融合与自动化标注。

2. 通过采集、标注、质量评估、训练数据交付完整闭环,把真实人类操作加工为可复用、可规模化交付训练资产;与元客视界整合硬件采集能力。

总3 招商合作与风向规避

1. 已与多家头部具身模型企业及机器人厂商项目级合作,输出高精度手部交互训练数据,沉淀标准化数据交付能力,可作为产业客户合作样本。

2. 多元资本联合入局提供资金与产业资源通路;需规避原始数据不可直接训练、工程化与商业化能力不足等风险。

总1 产业新动向

1. 具身智能从能力展示迈向真实任务落地,物理AI数据基础设施成为关键环节,手物交互高质量多模态数据被视为产业化关键短板。

2. 市场化VC、产业龙头与国资引导基金联合入局星忆,顺禧基金领投种子轮,凌云光领投种子+轮,中关村资本、一创投资跟投,体现资本对数据基础设施的布局。

总2 新问题与技术路径

1. 第一人称Ego数据并不缺少,但复杂手部操作涉及精细动作、频繁遮挡与复杂手物交互,原始数据难直接转化为有效训练数据。

2. 星忆以EgoKit和HBR Engine应对严重遮挡、图像畸变、高速运动、极端光照,做手部姿态解析、多模态融合、自动化标注,实现毫米级三维手部姿态解析与精细化标注。

总3 商业模式与政策启示

1. 商业模式是以算法为核心壁垒、数据质量为价值落点,提供采集、标注、质量评估、训练数据交付全链路数据基建服务,并通过凌云光·元客视界产投协同形成硬件+算法闭环。

2. 国资耐心资本支持硬科技成果产业化转化,多元资本协同为算法迭代和规模化生产打通资源通路;团队清华、北航背景及顶会论文80余篇构成研究背书。

返回默认

声明:快读内容全程由AI生成,请注意甄别信息。如您发现问题,请发送邮件至 run@ebrun.com 。

我是 品牌商 卖家 工厂 服务商 平台商 研究者 帮我再读一遍。

Quick Summary

Summary 1 Funding and company positioning

1. Xingyi Intelligence has completed consecutive seed and seed-plus rounds in the tens of millions of RMB. The seed round was led by Shunxi Fund; the seed-plus round was led by Luster LightTech, with participation from Zhongguancun Capital and Yichuang Investment.

2. The funds will focus on core algorithm and product R&D, real-world multimodal data infrastructure, and customer delivery and commercialization capabilities.

Summary 2 Data business highlights

1. Embodied AI is entering the task-deployment stage. Raw Ego data from complex hand operations is difficult to convert directly into training data; the bottleneck is high-quality, learnable, and scalably deliverable training data.

2. Using the EgoKit multimodal collection system and HBR Engine intelligent data engine, the company addresses severe occlusion, image distortion, high-speed motion, and extreme lighting through hand pose parsing, multimodal fusion processing, and automated annotation, turning real human operations into reusable training assets.

Summary 3 Partnerships and team

1. It has carried out project-level cooperation with several leading embodied model companies and robot manufacturers, delivering high-precision hand-interaction training data and building standardized data delivery capabilities.

2. Incubated at Tsinghua University's Department of Computer Science, the core team comes from Tsinghua and Beihang and combines robotics, AI, and autonomous driving industry experience, with more than 80 papers published at top conferences such as CVPR and NeurIPS.

Summary 1 Brand positioning and endorsement

1. Xingyi Intelligence positions itself as a physical AI data infrastructure provider for embodied AI. Incubated at Tsinghua University's Department of Computer Science, it builds a technology-driven brand label around algorithms as the core barrier and data quality as the value anchor.

2. The seed round was led by Shunxi Fund; the seed-plus round was led by Luster LightTech, with participation from Zhongguancun Capital and Yichuang Investment. Market-oriented VCs, industrial leaders, and state-guided funds jointly invested, endorsing the scarcity of its technology.

Summary 2 Product R&D and differentiation

1. It self-developed the EgoKit multimodal collection system and HBR Engine intelligent data engine, targeting four pain points—severe occlusion, image distortion, high-speed motion, and extreme lighting—to build hand pose parsing, multimodal fusion processing, and automated annotation capabilities.

2. For complex hand-object interaction, it processes real human operations into high-quality, reusable, and scalably deliverable training assets, building B2B differentiation through scarce data and standardized delivery.

Summary 3 Channels and industrial collaboration

1. Luster LightTech is both an industrial investor and a business collaboration partner. Through its wholly owned subsidiary Yuanke Vision, it has formed a strategic cooperation with Xingyi Intelligence, integrating first-person Ego collection, multi-view visual collection, and optical motion capture hardware capabilities.

2. It has project-level cooperation with several leading embodied model companies and robot manufacturers, delivering high-precision hand-interaction training data and building standardized data delivery capabilities, demonstrating synergy in industrial channels and delivery.

Summary 1 Market opportunity and growth signals

1. As embodied AI enters the task-deployment stage, high-quality training data for complex hand-object interaction is a scarce industry supply, and there is a large market gap for hand training data.

2. Xingyi focuses on first-person real interaction data and is building full-stack data infrastructure. It has project-level cooperation with several leading embodied model companies and robot manufacturers, indicating B2B demand and partnership opportunities.

Summary 2 Learnable business model and cooperation approach

1. With algorithms as the core barrier and data quality as the value anchor, it uses EgoKit collection, HBR Engine processing, hand pose parsing, multimodal fusion, and automated annotation to form a closed loop covering collection, annotation, quality assessment, and training data delivery.

2. Luster LightTech led the round as CVC industrial capital and formed strategic cooperation through Yuanke Vision, integrating hardware collection and automated data processing technology to create technology-plus-industrial-scenario synergy. State-owned patient capital supports the industrialization of hard-tech achievements.

Summary 3 Risks and response notes

1. Raw Ego data is not lacking, but complex hand operations involve fine movements, frequent occlusion, and complex hand-object interaction, making it difficult to convert directly into effective training data. High-quality, scalably deliverable data is the bottleneck.

2. The team needs to strengthen engineering implementation and commercialization capabilities. The article suggests that the essence of embodied AI competition is a contest in the supply capability for high-quality physical interaction data.

Summary 1 Product design and production needs

1. To address four technical pain points—severe occlusion, image distortion, high-speed motion, and extreme lighting—an algorithm system adapted to complex imaging and dynamic interaction scenarios is required.

2. The self-developed EgoKit multimodal collection system and HBR Engine intelligent data engine fuse visual, tactile, inertial, and other multi-source information, supporting millimeter-level 3D hand pose parsing and fine-grained annotation in long-sequence, high-contact, and human-object interaction scenarios.

Summary 2 Commercial opportunities

1. There is a market gap in hand training data for embodied AI. Xingyi has project-level cooperation with several leading embodied model companies and robot manufacturers, delivering high-precision hand-interaction training data.

2. Through Yuanke Vision, Luster LightTech has formed a strategic cooperation with Xingyi, integrating first-person Ego collection, multi-view visual collection, and optical motion capture hardware capabilities, bringing collaboration opportunities for high-end data collection matrices and engineering barriers.

Summary 3 Digitalization and e-commerce implications

1. Industrially processing real human operations into high-quality, reusable, and scalably deliverable training assets depends on automated annotation, multimodal fusion processing, quality assessment, and a standardized data delivery loop.

2. For manufacturing companies, the lesson is to connect scenario data to training assets through data collection equipment, algorithm processing, and standardized delivery.

Summary 1 Industry trends

1. Embodied AI has entered the task-deployment stage, and the industry is setting higher standards for real-world operation data. High-quality, learnable, and scalably deliverable training data has become a bottleneck on model capability ceilings.

2. Physical AI data infrastructure providers are emerging. Xingyi focuses on first-person real interaction data, with algorithms as the core barrier and data quality as the value anchor.

Summary 2 New technologies and solutions

1. The EgoKit multimodal collection system and HBR Engine intelligent data engine provide hand pose parsing, multimodal fusion processing, and automated annotation capabilities for severe occlusion, image distortion, high-speed motion, and extreme lighting.

2. They process real human operations into high-quality, reusable, and scalably deliverable training assets, forming a complete business loop from collection and annotation to quality assessment and training data delivery.

Summary 3 Customer pain points and service opportunities

1. The customer pain point is that raw Ego data is abundant, but data on complex fine hand movements, frequent occlusion, and complex hand-object interaction is difficult to use directly for training. What is scarce is high-quality data that can be delivered at scale.

2. Xingyi has project-level cooperation with several leading embodied model companies and robot manufacturers, delivering high-precision hand-interaction training data and building standardized data delivery capabilities. Hardware collaboration with Yuanke Vision can enhance its service capabilities.

Summary 1 Platform requirements and problems

1. Task deployment in embodied AI requires real-world operation data, but raw Ego data is difficult to convert directly into training data. High-quality, learnable, and scalably deliverable training data is the bottleneck.

2. The platform needs to solve data processing problems in complex hand operations, including fine movements, frequent occlusion, and complex hand-object interaction, and to establish quality assessment and standardized delivery capabilities.

Summary 2 Latest practices and operational management

1. With algorithms as the core barrier, Xingyi self-developed the EgoKit multimodal collection system and HBR Engine intelligent data engine to perform hand pose parsing, multimodal fusion, and automated annotation for the four technical pain points.

2. Through a complete loop covering collection, annotation, quality assessment, and training data delivery, it processes real human operations into reusable and scalably deliverable training assets. It also integrates hardware collection capabilities with Yuanke Vision.

Summary 3 Partnership recruitment and risk avoidance

1. It has project-level cooperation with several leading embodied model companies and robot manufacturers, delivering high-precision hand-interaction training data and building standardized data delivery capabilities, which can serve as a sample for industrial customer cooperation.

2. Diversified capital participation provides funding and industrial resource channels. Risks to avoid include raw data that cannot be directly used for training and insufficient engineering and commercialization capabilities.

Summary 1 New industry developments

1. Embodied AI is moving from capability demonstrations to real task deployment, and physical AI data infrastructure has become a key link. High-quality multimodal data for hand-object interaction is regarded as a key industrial shortcoming.

2. Market-oriented VCs, industrial leaders, and state-guided funds jointly invested in Xingyi. Shunxi Fund led the seed round, Luster LightTech led the seed-plus round, and Zhongguancun Capital and Yichuang Investment participated, reflecting capital's positioning in data infrastructure.

Summary 2 New problems and technical paths

1. First-person Ego data is not lacking, but complex hand operations involve fine movements, frequent occlusion, and complex hand-object interaction, so raw data is difficult to convert directly into effective training data.

2. Xingyi uses EgoKit and HBR Engine to address severe occlusion, image distortion, high-speed motion, and extreme lighting through hand pose parsing, multimodal fusion, and automated annotation, achieving millimeter-level 3D hand pose parsing and fine-grained annotation.

Summary 3 Business model and policy implications

1. The business model uses algorithms as the core barrier and data quality as the value anchor, providing full-chain data infrastructure services covering collection, annotation, quality assessment, and training data delivery. Through the industrial-investment collaboration with Luster LightTech and Yuanke Vision, it forms a hardware-plus-algorithm loop.

2. State-owned patient capital supports the industrialization of hard-tech achievements, and diversified capital collaboration opens resource channels for algorithm iteration and scaled production. The team's Tsinghua and Beihang background and more than 80 top-conference papers provide research endorsement.

Disclaimer: The "Quick Summary" content is entirely generated by AI. Please exercise discretion when interpreting the information. For issues or corrections, please email run@ebrun.com .

I am a Brand Seller Factory Service Provider Marketplace Seller Researcher Read it again.

星忆智能完成数千万元种子轮及种子+轮连续融资。种子轮由顺禧基金领投,种子+轮由产业资本凌云光领投,中关村资本、一创投资跟投。资本与产业双向加持,加速高质量手物交互数据规模化交付。

近日,星忆智能宣布完成数千万元种子轮及种子+轮连续融资。其中种子轮由顺禧基金领投;种子 + 轮由产业资本凌云光领投,中关村资本、一创投资跟投。本轮融资集合市场化VC、产业龙头与国资引导基金等多元资本力量,资金将重点投入核心算法与产品研发、真实世界多模态数据基础设施建设,以及客户交付与商业化能力提升。

图片

具身智能迈入任务落地阶段,行业对真实世界操作数据提出更高标准。市场上并不缺少第一人称Ego数据,但复杂的手部操作涉及精细动作、频繁遮挡与复杂手物交互,原始数据难以直接转化为有效的训练数据。高质量、可学习、可规模化交付的训练数据,才是制约模型能力上限的真正瓶颈。

图片

星忆智能围绕核心算法构建全栈数据基础设施体系

星忆智能孵化于清华大学计算机系,致力于打造全球领先的物理AI数据基础设施,以算法为核心壁垒,以数据质量为价值落点,聚焦第一人称真实交互数据。依托自研EgoKit多模态采集系统与HBR Engine智能数据引擎,针对严重遮挡、图像畸变、高速运动、极端光照四大技术痛点,搭建适配复杂成像与动态交互场景的算法体系;通过手部姿态解析、多模态融合处理与自动化标注能力,把真实人类操作,加工为高质量、可复用、可规模化交付的训练资产。

图片

星忆智能核心算法亮点

图片

本轮融资中,凌云光兼具产业投资方与业务协同伙伴双重身份:既是种子+轮领投方,同时通过旗下全资子公司元客视界与星忆智能达成深度战略合作。双方锚定具身智能爆发带来的手部训练数据巨大市场缺口,整合第一视角Ego采集、多视点视觉采集、光学动捕硬件能力与星忆智能自动化数据处理技术,共建覆盖采集、标注、质量评估、训练数据交付的完整业务闭环。

作为CVC产业资本,凌云光以领投实现资本与业务深度联动,既为星忆扩充高端数据采集矩阵、夯实技术与工程壁垒注入资源,也验证了 “技术能力+产业场景” 双向认可的协同模式。目前,星忆智能已与多家头部具身模型企业及机器人厂商开展项目级合作,输出高精度手部交互训练数据,沉淀标准化的数据交付能力。

图片

面对具身智能产业化浪潮,市场化机构、产业龙头基于对赛道与技术的深度研判,给出了各自的判断。

种子轮领投方顺禧基金表示:

“具身智能迈向真实任务落地,复杂手物交互的高质量训练数据属于行业稀缺供给。顺禧基金作为北京国资背景的耐心资本,重点支持硬科技成果产业化转化。我们非常看重星忆智能的底层算法壁垒,其攻克复杂工况下手部姿态解析难题,具备将原始场景数据工业化加工为训练资产的全链路能力,拥有数据规模化交付的发展潜力。我们期待陪伴团队持续迭代,助力国内具身智能底层数据基础设施能力建设。”

种子+轮领投方凌云光表示:

“凌云光长期布局工业AI与具身智能赛道,此前也战略投资智谱华章等前沿AI项目。基于我们多年工业场景落地的实践视角,我们对技术的稀缺性、工程化能力与市场化潜力有着严苛的判断标准。手物交互高质量多模态数据是具身智能产业化的关键短板,星忆智能在复杂工况手部数据解析上具备稀缺的算法能力,同时拥有从算法到交付的完整工程化能力。我们看好团队把技术转化为规模化数据供给的发展潜力,也期待依托双方硬件与算法协同,共同推动具身智能数据基础设施走向产业落地。”

图片

市场化机构、产业龙头与国资平台的联合入局,是行业对星忆智能核心技术稀缺性的有力背书。多元资本形成协同合力,一方面为底层算法的持续迭代提供资金支撑,另一方面为技术落地数据规模化生产打通产业资源通路,推动公司从单点技术优势,进阶为具备完整能力的数据基建服务商。

星忆智能创始人兼CEO宋知珩表示:

非常感谢顺禧基金、凌云光及全体投资方的认可。国资耐心资本与工业产业龙头的双重加持,既是对我们手部交互多模态数据处理稀缺技术能力的肯定,也对团队的工程落地与商业化能力提出更高要求。具身智能赛道的竞争,本质是高质量物理交互数据供给能力的比拼。我们将持续深耕算法底座,依托与凌云光・元客视界的产投协同,打通从硬件采集到标准化数据集交付的全链路,将实验室算法能力转化为稳定、可规模化的数据基建服务,赋能模型与机器人企业,推动具身智能技术真正落地真实场景。

关于星忆智能

星忆智能(ASTRALIVE)孵化于清华大学计算机系,致力于成为全球领先的具身智能物理AI数据基础设施服务商。公司以高精度手部姿态理解与数据标注技术为核心,自研多模态数据采集设备与数据处理引擎,融合视觉、触觉、惯性等多源信息,攻克遮挡、强光、高速运动等真实场景难题,实现长序列、强接触及人手—物体交互场景下的毫米级三维手部姿态解析与精细化标注。核心团队源自清华、北航,兼具机器人、AI、自动驾驶汽车头部产业经验,累计发表CVPR、NeurIPS等顶会论文80余篇,斩获Oral、Highlight、Outstanding Paper等多项学术荣誉。

注:文/龚作仁,文章来源:Laborer,本文为作者独立观点,不代表亿邦动力立场。

文章来源:Laborer

广告
微信
朋友圈

FAQ回顾

星忆智能是一家什么公司?

星忆智能(ASTRALIVE)孵化于清华大学计算机系,定位为具身智能物理AI数据基础设施服务商,聚焦第一人称真实交互数据。公司自研EgoKit多模态采集系统与HBR Engine智能数据引擎,提供面向复杂手物交互的高精度手部交互训练数据与标准化数据交付能力。

具身智能为什么需要高质量手部交互训练数据?

具身智能进入真实任务落地阶段后,复杂手部操作涉及精细动作、频繁遮挡和复杂手物交互,原始第一人称Ego数据往往难以直接转化为有效训练数据。高质量、可学习、可规模化交付的训练数据,是制约模型能力上限的关键瓶颈。

星忆智能的核心技术能解决哪些数据采集难题?

星忆智能围绕严重遮挡、图像畸变、高速运动和极端光照四类痛点搭建算法体系,具备手部姿态解析、多模态融合处理和自动化标注能力,可将真实人类操作加工为高质量、可复用、可规模化交付的训练资产,并实现毫米级三维手部姿态解析与精细化标注。

星忆智能与凌云光、元客视界的合作内容是什么?

凌云光在种子+轮领投星忆智能,并通过全资子公司元客视界与星忆智能达成战略合作。双方整合第一视角Ego采集、多视点视觉采集、光学动捕硬件能力与星忆智能自动化数据处理技术,共建覆盖采集、标注、质量评估、训练数据交付的完整业务闭环。

具身智能数据基础设施服务商主要服务哪些客户?

星忆智能主要面向具身模型企业和机器人厂商,提供高精度手部交互训练数据。公司已与多家头部具身模型企业及机器人厂商开展项目级合作,并沉淀标准化数据交付能力,帮助客户获得可规模化使用的真实世界多模态训练数据。

这么好看,分享一下?

朋友圈 分享

APP内打开

赞 +1
+1
微信好友 朋友圈 新浪微博 QQ空间
关闭
收藏成功
发送
/140 0