Back to List

IO-AI Open Source ROSView: An Offline Visualization Tool for Embodied Intelligence Data

Over the past few years, Foxglove Studio has been one of the visualization tools widely used by global robotics developers. However, with its full transition to closed source by the end of 2024, the robotics community is facing challenges such as limited toolchains and difficulties in secondary development.

In the current era of rapid development of embodied intelligence, the industry urgently needs a truly open, freely extensible, and deeply compatible robot data visualization platform.

IO-AI TECH officially announced the open-sourcing of ROSView, a browser-based robot data visualization platform specifically designed for robot and embodied intelligence R&D scenarios. It supports multiple robot data formats, features rich visualization panels, and has high-performance playback capabilities for large-scale data, aiming to provide developers in the robotics industry with an easy-to-use, open, and sustainably evolving foundational tool.

Multi-format deep compatibility: Breaking down data silos

ROSView covers mainstream data formats in the fields of various robots and embodied intelligence, requiring no transcoding and being ready to use right away:

  • MCAP: The new generation of recording format officially recommended by ROS

  • ROS 1 bag: Compatible with ROS 1 historical recorded data

  • ROS 2 db3: Supports the default SQLite recording format of ROS 2

  • HDF5 / h5: Suitable for scientific research, MultiModal Machine Learning datasets, and AI training data scenarios

  • BVH: Supports viewing skeletal animation and motion capture data

All data can be directly parsed and visualized on thebrowser side,without the need to upload to a serverfor transcoding.

This means that developers can directly open data files locally, quickly complete debugging, analysis, and playback. This not only ** reduces the time cost of uploading large files**, but also fundamentally ensures the ** privacy and security** of core R&D data.

Performance optimization for large-scale robot data

Robot data typically includes high-frequency images, point clouds, TF, joint states, and various sensor data, placing extremely high demands on browser performance and data links.

ROSView has been specifically optimized for this scenario in its underlying architecture.

1. SharedArrayBuffer High-Performance Data Path

ROSView preferentially uses SharedArrayBuffer circular buffer technology to transfer large chunks of binary data between the Worker and the main thread via shared memory references.

For high-throughput data such as point clouds and images, it can significantly reduce memory copy and serialization overhead.

In browser environments that do not meet the SharedArrayBuffer conditions, the system will also automatically degrade to a compatible solution to ensure stable operation.

Even when dealing with multi-modal robot data files of several gigabytes, it can still maintain a stable and smooth loading and playback experience.

2. Worker + WebCodecs Image Decoding Pipeline

ROSView migrates all image decoding operations to the Worker for execution and combines with WebCodecs to perform hardware-accelerated processing on the video stream.

This means that even when playing simultaneously:

  • High Frequency Image Stream

  • Point Cloud Data

  • Joint Status

  • TF Data

  • Raw Message

The main interface can still maintain smooth responsiveness.

For high-bandwidth scenarios such as H.264 video streams, a better real-time playback experience can also be achieved.

3. Data Quality Scanning Capability

ROSView has built-in data quality scanning capabilities, which can automatically detect common issues in robot data, including:

  • Timestamp Rollback

  • Time Drift Anomaly

  • Frame Loss

  • Message time gap is abnormal

  • Topic data is discontinuous

Helps R&D personnel quickly locate issues in Data Acquisition and Recording.

4. Can be used independently or embedded and integrated

ROSView can either be used directly as an independent web page or embedded as a React component into enterprise internal platforms or custom systems.

Flexible Visualization Panel: Designed for Robot Debugging

ROSView provides a flexible multi-panel layout system that supports dragging, docking, and combination, making it convenient for R&D personnel to view data from multiple dimensions simultaneously.

Currently supported typical visualization capabilities include:

  • Image and Video Stream Viewing (Supports H.264 Image Data)

  • 3D Scene, Point Cloud, URDF, TF Visualization

  • Plot Curve and Time Series Analysis

  • Joint / Joint Status View

  • Map、Pose、Topic Graph、Raw Messages

  • Viewing Audio and Other MultiModal Machine Learning Data

  • Playback capabilities such as play, pause, frame-by-frame forward, speed control, etc.

R&D personnel can complete the entire debugging process from "detecting anomalies" to "locating the cause" on the same page:

First, view the image and point cloud, then align the time axis to analyze joint, TF, and Topic messages, and finally return to the raw data to confirm the source of the problem.

Permissive License: Supports commercial use and secondary development

ROSView is open source under the MIT License.

This means you can:

  • Directly use ROSView in commercial projects

  • Secondary development based on ROSView

  • Integrate ROSView into robot platforms, data platforms, or internal tools

  • Self-deploy, modify, and extend functionality

You only need to retain the copyright and license notice as required by the MIT License, and there is no need to apply for authorization.

Two Usage Modes

Direct Online Use

Open:https://rosview.com

can directly load and view robot data in the browser.

Embed as a React component into an enterprise internal platform or custom system

ROSView has been published as an npm package and can be directly embedded into React applications, integrated into:

  • Enterprise Internal Data Platform

  • Data Annotation System

  • Collection and Quality Inspection Platform

  • Model Training Data Platform

  • Robot Operation and Maintenance System

and other arbitrary scenarios.

Join the open-source ecosystem

AIO Intelligence has long been deeply involved in the data infrastructure of robotics and embodied intelligence, accumulating rich experience in large-scale data operations and technology. ROSView is an important practice of ours in the open-source ecosystem.

We hope that ROSView can serve more robot R&D teams, and we also welcome community developers to jointly participate in the construction:

  • Try it out and provide feedback on issues

  • Submit new format support

  • Add more visualization panels

  • Optimize performance for large files

  • Jointly Build the Robot Data Tool Ecosystem

GitHub Project Address:https://github.com/ioai-tech/rosview