Rerun.io

Rerun.io is an open-source tool designed to provide real-time visual debugging for computer vision (CV), robotics, and AI-driven perception systems. It allows developers to stream structured data—such as images, 3D point clouds, bounding boxes, and sensor logs—into a unified interface, enabling better insight, faster iteration, and more effective problem-solving.

Traditionally, debugging CV and robotics systems involves complex logging, manual playback, or disconnected visualizations. Rerun.io changes that by providing a powerful SDK and UI that makes debugging as intuitive as watching your system unfold in real time.

Whether you’re training an object detection model, working on autonomous navigation, or integrating a perception pipeline into a robot, Rerun.io delivers the tooling to see what your AI sees.

Features

Rerun.io includes a rich feature set tailored to the needs of CV and robotics developers:

• Real-Time Streaming of Visual Data
  Stream and visualize logs like images, tensors, 3D point clouds, and segmentation masks as your model runs.

• Time-Synchronized Data Views
  Inspect multiple data modalities (e.g., camera frames, depth maps, telemetry) on a synchronized timeline.

• Python & Rust SDKs
  Log data easily from Python or Rust-based applications, including deep learning and robotics pipelines.

• Interactive 2D/3D Visualization
  Explore image sequences, point clouds, coordinate frames, and annotated bounding boxes in real-time.

• Multi-Sensor Support
  Visualize data from multiple sensors (camera, LiDAR, radar, etc.) and align them spatially and temporally.

• Efficient Logging Format
  Lightweight and fast serialization optimized for high-throughput logging of dense visual data.

• Cross-Platform Viewer
  Use the Rerun viewer on Linux, macOS, or Windows to explore your logs interactively.

• Open Source & Extensible
  Built with open standards and available under the permissive MIT license for easy integration and contribution.

How It Works

Rerun.io fits into the development and debugging loop of perception-based systems in a few simple steps:

1. Integrate the SDK
   Install the Rerun SDK in your Python or Rust project. Set up logging calls in your pipeline to stream relevant data.

2. Log Structured Data
   Log video frames, sensor inputs, detection results, object poses, and more using the intuitive API.

3. Run and Stream
   Execute your application. Data is streamed in real time to the Rerun viewer or recorded for playback.

4. Explore in Viewer
   Use the UI to inspect timelines, 3D scenes, and sensor overlays. Filter by tags or component for focused debugging.

5. Iterate and Improve
   Adjust your model or system with insights gained from visual inspection and re-run the workflow.

Use Cases

Rerun.io is designed for a wide array of perception and robotics development tasks:

• Computer Vision Model Debugging
  Inspect detection errors, segmentation overlays, or pose estimation in real time.

• Robot Perception Systems
  Visualize how your robot interprets the world through its sensors and vision stack.

• Simulated Environment Testing
  Stream synthetic data from simulators into Rerun to debug perception logic before hardware deployment.

• SLAM and Mapping Projects
  View keypoints, trajectories, and environmental features during simultaneous localization and mapping.

• Autonomous Driving Prototypes
  Overlay bounding boxes and LIDAR data with video feeds for comprehensive system validation.

• Edge AI Deployment
  Log and inspect model outputs from edge devices like Jetson, Raspberry Pi, or robotics boards.

Pricing

Rerun.io is currently open-source and free to use.

• Open Source (MIT License)

  • Free for personal, academic, and commercial use

  • Full access to all features

  • Source code available on GitHub

  • Community support via GitHub Discussions and Discord

• Future Commercial Offerings (Planned)

  • Hosted data dashboards

  • Team collaboration features

  • Cloud log storage and sharing

  • Enterprise integrations

As of now, Rerun is fully functional and free for local use, with premium enterprise tools in development.

Strengths

Rerun.io offers a number of compelling advantages for AI and robotics developers:

• Purpose-Built for Perception Systems
  Tailored to the data formats and workflows used in CV and robotics applications.

• Real-Time and Historical Views
  Analyze live data streams or recorded sessions with equal ease.

• Developer-Friendly SDKs
  Simple, intuitive APIs for Python and Rust accelerate adoption.

• Open and Transparent
  Fully open source with an active community and transparent roadmap.

• Multi-Modal Data Support
  See multiple layers of data—images, annotations, and spatial transforms—in one unified UI.

• Efficient Performance
  Built with performance in mind to support high-frequency, large-scale logging.

Drawbacks

While Rerun.io is powerful, there are some limitations to be aware of:

• Not a Full ML Ops Tool
  Rerun is focused on visualization and debugging—not model training, versioning, or deployment.

• Early Stage for Enterprise Features
  Team dashboards, cloud storage, and collaboration tools are not yet generally available.

• Limited Language Support
  Currently supports Python and Rust; no native support for C++, JavaScript, or other languages.

• Manual Setup Required
  Though lightweight, initial SDK integration requires some setup in your existing codebase.

Comparison with Other Tools

Here’s how Rerun.io stacks up against similar debugging or visualization tools:

• Versus TensorBoard
  TensorBoard is great for training metrics. Rerun focuses on real-time visual perception debugging with 2D/3D views.

• Versus RViz (ROS)
  RViz is tightly coupled with ROS. Rerun offers broader integration without requiring ROS infrastructure.

• Versus FiftyOne
  FiftyOne is ideal for dataset analysis. Rerun is better for live and interactive debugging during model execution.

• Versus Custom Visualization Tools
  Many teams build internal tools for visual logging. Rerun eliminates the need with a plug-and-play viewer.

Rerun fills a critical gap between raw sensor data and deep learning model outputs—providing real-time visual feedback to accelerate iteration.

Customer Reviews and Testimonials

Rerun.io has received strong praise from early users in the computer vision and robotics space:

• “This is the tool I’ve always wanted but never had time to build.” – Robotics Engineer

• “Rerun helped us catch a calibration bug that took hours to spot in logs before.” – CV Researcher

• “It’s like TensorBoard, but for actual perception systems—awesome work.” – AI Developer

• “Our debugging workflow went from guesswork to guided insights.” – Autonomous Systems Engineer

The open-source community has also embraced Rerun with active contributions and feedback on GitHub.

Conclusion

Rerun.io is a powerful, open-source tool purpose-built for developers working on perception-heavy systems such as computer vision pipelines and robotics platforms. With real-time visual debugging, multi-modal support, and intuitive integration, Rerun shortens the feedback loop between development and insight.

If you’re building intelligent systems that “see” the world, Rerun.io is the fastest way to see what your model sees—and fix what it doesn’t.