IO-AI Data Platform
IO-AI Data Platform provides end-to-end full-chain capabilities from data to models, helping teams quickly establish sustainable data closed loops and delivery processes.
Software Advantages
"Universal" Robot Platform
The platform is based on ROS and provides pluggable preprocessing/postprocessing capabilities for devices of different configurations and vendors, reducing the cost of data governance across robots and formats.
Multiple robot acquisition systems have been adapted (such as Zhiyuan, Songling, Ruierman, etc.). Different format data can be automatically converted to standard ROS standards for unified management, annotation, export, and training.
Know-How Accumulation
The team has accumulated extensive experience in data collection and model training, having processed and annotated thousands of hours of multimodal data, and productized the methodological practices into standard platform functions to improve delivery efficiency and consistency.
In addition to data processes, the platform also supports centralized management of personnel and equipment, such as unified monitoring and take-over of robot devices:
Seamless Integration with IO Products
The platform integrates deeply with IO's full-body motion capture SenseXperience and teleoperation TeleXperience, achieving seamless connection from acquisition to streaming, training.
Purchase IO's universal robot teleoperation product to receive the IO Data Platform experience version.
Data Security Guarantee
The platform provides enterprise-level security and compliance capabilities: support for private offline deployment, IP whitelists, access/operation audit logs, etc., covering key risk points.
Adopts a storage and program separation architecture: data can be stored in customer-owned object storage, with access permissions completely controlled by the customer.
Flexible Personalization Customization
Support customization according to enterprise needs, quickly create exclusive platform forms:
- Brand Visual Customization: Customize Logo, Favicon, title and introduction, etc.
- Theme Style Customization: Upload and manage custom themes, preview effects in real-time.
- Robot Adaptation Support: Convert any configuration and format data uniformly to standard ROS standards for centralized management.
Deployment Methods
1. SaaS Cloud Deployment
- Ready-to-use: No local installation required, access the platform via the internet after opening an account.
- Automatic Updates: Continuously obtain the latest functions and security updates.
- Elastic Scaling: Allocate resources on demand to adapt to projects of different scales.
- Delivery Contents: Dedicated subdomain, initial accounts and passwords for each role.
Delivery Checklist:
- Platform access dedicated subdomain
- Initial accounts and passwords for each role
- User operation manual and usage documentation
- Technical support and service contact information
2. Private Mirror Deployment
- Data Localization: The platform can be deployed on customer-owned servers or private clouds, with complete customer control over data.
- Customized Development: Functional customization and system integration can be performed according to business needs.
- High Availability: Independent operation ensures business continuity and data security.
- Delivery Contents: Docker image offline package, on-site support services, optional data storage arrays.
Delivery Checklist:
- Platform Docker image offline installation package
- Installation and deployment description documents
- On-site deployment and technical support services
- Data storage array servers (optional, if purchased)
- User operation manual and maintenance documents
Both deployment methods provide regular upgrade services (such as new export formats, etc.), subject to the warranty period agreed in the contract.
Core Functions Overview
| Function Module | Main Functions | Goals and Values |
|---|---|---|
| User and Permission Management | Accounts and roles, permission control, IP whitelist, access audit | Strengthen security and compliance to meet enterprise-level requirements |
| Project and Task Management | Project configuration, task assignment, progress monitoring, statistical analysis, collaboration and queuing management | Improve collaboration efficiency and process transparency |
| Data Collection and Annotation | Multi-source collection (video/image/sensor), annotation tools, process automation, batch annotation | Improve efficiency and consistency, covering diverse data sources |
| Data Preprocessing and Postprocessing | Visualization, format conversion (video/audio to MCAP), quality inspection, standardization and compatibility detection | Ensure data quality and improve training availability |
| Review and Feedback | Annotation review, return for rework, quality assessment and closed-loop feedback | Maintain quality standards and continuously optimize processes |
| Report Generation | Progress and quality report export, automated report generation | Support management decisions and facilitate external presentation |
| Data Import | Cloud storage docking, batch import, multi-source access, soft delete recovery | Rapid access and governance, reduce migration costs |
| Data Export | Annotated result export, LeRobot/HDF5/MCAP and other formats, export history and progress tracking | Convenient streaming, adapt to multi-training/evaluation chains |
| Model Training | PyTorch/JAX, multi-platform (CUDA/MPS/CPU), parameter configuration, monitoring and checkpoint management | Unify training closed-loop, improve experiment reproducibility and controllability |
| Model Inference | One-click deployment, simulation testing, MCAP testing, offline edge deployment and service management | End-to-end inference delivery from verification to online |
Permissions and Data Security
- Role System: Administrators, project managers, auditors, annotators, collectors; authorize on demand, following the principle of least privilege.
- Project Isolation: Isolate data and users by project; members can only access data within the scope of assigned projects.
- Data Sovereignty: The platform can only save data access links; data is actually stored in customer-owned or designated cloud storage.
- Security Mechanisms: IP whitelist, access keys, encrypted transmission, fine-grained permissions and audit logs, etc.
Contact and Support
If you need customized solutions or platform trials, please contact us:
- Email: io@io-ai.tech
- Phone: (+86) 0755-88658665
- Address: Room 5, Floor 1, Building C1, Phoenix Wisdom City, Phoenix Street, Guangming District, Shenzhen
- Website: https://io-ai.tech
Welcome to register and experience the free open IO Data Open Platform, understand platform functions from the project manager perspective.