Rerun

Rerun / The Physical Data Platform Iterate faster on robotics learning with unified infrastructure. Ingest, visualize, annotate, query, and transform - from collection to training. Try browser demo Explore examples Rerun is open source 6,200 Modern infrastructure for robotics learning Try browser demo Quick start guide Visualize Visualize consistently as you explore, build, evaluate, and debug Manage and share all data and visualizations in one place Explore examples Query & Transform Run snappy dataframe queries optimized for robotics data Transform logs into training-ready data with simple pipelines Read docs Ingest, Store & Retrieve Ingest, store, and index robotics logs of multi-formats Search and retrieve petabyte-scale robotics data in seconds Book a demo Start visualizing in seconds Powerful and flexible visualization for spatial and embodied AI that's shockingly easy to get started with. C++PythonRust pip install rerun-sdk rerun Python quick start guide Try browser demo The community loves building on Rerun LeRobot Brush PyCuVSLAM Ultra Project Aria LeRobot LeRobot is Huggingface's State-of-the-art AI for real-world robotics project. They are using Rerun as an integrated part of their visualization tools. Go to project → LeRobot LeRobot is Huggingface's State-of-the-art AI for real-world robotics project. They are using Rerun as an integrated part of their visualization tools. Go to project → Brush Brush is a 3D reconstruction engine using Gaussian splatting developed by Arthur Brussee at Deepmind. It is written in Rust and aims to be highly portable, flexible and fast. Rerun is used for visualization during training. Go to project → PyCuVSLAM PyCuVSLAM is the official Python wrapper for NVIDIA's cuVSLAM library, providing GPU-accelerated visual SLAM and camera tracking for real-time localization and mapping. Go to project → Ultra Ultra is building intelligent warehouse robots that automate repetitive, dangerous, and variable tasks. They use Rerun Data Platform for their end-to-end data visualization workflow and transformation pipeline. Go to project → Project Aria Project Aria is a research platform developed by Meta Reality Labs Research to push the state of the art in egocentric AI research. Rerun is used to visualize sequences in their Aria Dataset Explorer. Go to project → The fastest way to build Physical AI Rerun accelerates your speed of progress by helping you SEE faster Spot Spot interesting moments from robot logs that could improve your model - a failure in a perception pipeline, an unexpected action, or a novel real-world scenario.Python # Find rows around the time of the anomaly around_anomaly = dataset.filter( dn.Expr.between( col("time"), anomaly_time - 10, anomaly_time + 10 ) ) Expand Zoom out from a single moment to the surrounding context to trace root cause.Find all similar events across your recordings to collect training data. Experiment Transform messy logs into training-ready datasets with a consistent and flexible pipeline.Debug visually as you train, evaluate, and iterate ML models. Rerun Open Source Visualize robot runs and turn them into training data locally. Try browser demo Read the documentation Log & Model Multi-format robotics log ingestionUnified data model for both recording and tabular dataData modeling simple by default, flexible by design C++PythonRust import rerun as rr rr.init("my_data_generating_application") rr.connect() # Connect to a remote viewer … rr.log("points", rr.Points3D(positions)) rr.log("camera", rr.Transform3D(pos, rot)) rr.log("camera/image", rr.Pinhole(intrinsics)) rr.log("camera/image", rr.Image(tensor)) rr.log("reprojection_error", rr.Scalar(err))View & Customize The fastest, most powerful visualizer for Physical AIRun it natively and in the browser; or use it to power your custom toolsCustom layouts and visualizations via UI or code Analyze & Prepare Powerful dataframe queries for filtering and analysisBuilt-in robotics operations simplifying data pipelineThe fastest path from messy logs to training data Rerun Data Platform Built on open source, extended for production. Consistently produce large-scale, high-quality training data from massive robot runs. Book a demo See comparison Manage centrally, search instantly The fastest indexing, search, and retrieval engine for robotics dataAll your messy and clean data organized in one placeVersion, govern, and discover easily - no single detail lost Transform consistently, scale effortlessly Purpose-built query engine for complex, large-scale robotics dataDefine only once, transform incrementally and consistentlySchema evolution with fewer pipeline interruptions Python # Find the latest embedding for the selected time embedding_table = ( dataset.filter_contents( contents="/camera/wrist/embedding", ).reader( index="real_time", ).limit(1) ) embedding = ( embedding_table .collect_column("/camera/wrist/embedding:embeddings")[0] )Share easily, debug collaboratively One-click sharing of data, snapshots, and visualizationsCollaborate easily with shared visual context across your entire workflow Enterprise-ready Physical AI Enterprise SSO and security controls to safeguard your data and IPDeployed in your cloud and region of choice; data stored in your own bucket Build intelligence not infrastructure Winning teams win on the speed of product iteration. Rerun scales your infrastructure and simplifies your data pipelines. Every engineering hour drives new capability, not plumbing Rerun Open Source vs. Rerun Data Platform Feature Comparison Rerun Open Source Visualize robot runs and turn them into training datasets, run locally Dual licensed under MIT and Apache 2 Join on GitHub Robotics log ingestion Data modeling Dataframe queries for robotics data UDF support Single file sharing Simple in-memory data catalog Rerun Data Platform Turn large scale robotics log data into high-quality training data Book a demo Book a demo All Rerun Open Source features plus: Centralized data catalog for large-scale robotics data Centralized data management (versioning, discovery, lineage) Managed cloud platform for scaled data infrastructure Indexing, search, and retrieval Optimized query engine for speed and scale Incremental data transformation and schema evolution Sharing and collaboration Enterprise security and compliance Rerun Open Source Visualize robot runs and turn them into training datasets, run locally Dual licensed under MIT and Apache 2 Join Github Rerun Data Platform Turn large scale robotics log data into high-quality training data Beta release with selected customers Book a demo Robotics log ingestion ✓ ✓Data modeling ✓ ✓Data visualization ✓ ✓Dataframe queries for robotics data ✓ ✓UDF support ✓ ✓Single file sharing ✓ ✓Data catalog ✓ Simple in-memory catalog ✓ Centralized catalog for large-scale robotics dataCentralized data management (versioning, discovery, lineage) ✕ ✓Managed cloud platform for scaled data infrastructure ✕ ✓Indexing, search, and retrieval ✕ ✓Optimized query engine for speed and scale ✕ ✓Incremental data transformation and schema evolution ✕ ✓Sharing and collaboration ✕ ✓Enterprise security and compliance ✕ ✓ Join Github Book a demo Our blog Rerun 0.27 - Flexible transforms, Python server management, and improved time controls Rerun 0.27 includes experimental coordinate frame hierarchies, Python APIs for server management, blueprint controls for 3D views, and time panel improvements. Rerun 0.26 - ROS2 reflection, transform performance and more Rerun 0.26 brings major performance improvements, reflection-based ROS2 MCAP support, experimental lenses, and continued polish across the viewer and SDK. Sharing in Rerun - from web to native viewer Learn how Rerun's URL-based architecture and open source design make sharing Physical AI visualizations effortless—from command line to web viewer to native apps. Browse all blog posts GitHub Discord LinkedIn Twitter DocsBlogExamplesTeamCareers The Rust logo is a trademark owned by The Rust foundation Privacy policy --- Rerun / GitHub Discord LinkedIn Twitter DocsBlogExamplesTeamCareers Privacy policy --- Rerun / Examples Spatial computing and XR Examples related to spatial computing, augmented reality, virtual reality, and mixed reality. View ARKit scenes 2D3DDepthMeshObject detectionPinhole cameraBlueprint Source code 3D line mapping revisited 2D3DStructure from motionTime seriesLine detectionPinhole cameraPaper walkthrough Source code Objectron 2D3DObject detectionPinhole cameraBlueprint Source code SimpleRecon: 3D reconstruction without 3D convolutions 3DDepthTime seriesPinhole cameraMeshPaper walkthrough Source code Decoupling human and camera motion from videos in the wild 3DSLAMKeypoint detectionMeshTime seriesPaper walkthrough Source code VRS viewer 2D3DVRSViewerC++ Source code Learning to render novel views from wide-baseline stereo pairs 2D3DView synthesisTime seriesPinhole cameraPaper walkthrough Source code ARFlow: a framework for simplifying AR experimentation workflow 3DAugmented realitySpatial computingIntegration Source codeRobotics Examples related to robotics, autonomous systems, and interfacing with sensor hardware. View IMU signals Plots Source codeView RGBD 2D3DDepthNYUDPinhole camera Source codeView RRT* 2D Source codeView nuScenes Lidar3D2DObject detectionPinhole cameraBlueprint Source codeView URDF 3DMeshURDFAnimation Source code DROID 2D3DDepthPinhole cameraBlueprint Source code LeRobot dataset from RRD RoboticsMCAPLeRobotDatasetServer Source code ROS node 2D3DPinhole cameraROSTime seriesURDF Source code Robby fischer 3DURDFBlueprint Source code ROS 2 bridge 2D3DPinhole cameraROSTime seriesC++ Source code ROS bridge 2D3DMeshPinhole cameraROSTime seriesC++ Source code KISS-ICP 3DPoint cloudLidar Source code Live depth sensor 2D3DLiveDepthRealSense Source code Lidar Lidar3D Source code NV12 2DImage encodingYUV Source code MCAP MCAPRRDROSROS 2RosbagTutorial Source code Eye control Eye control3DPinhole camera Source code ROS TF ROSTFTransformCoordinate FrameROS 2 Source code Any scalar Any scalarPlottingDynamicArchetype Source codeDiffusion models, LLMs, and machine learning Examples using machine learning and generative AI methods such as diffusion and LLMs. ControlNet ControlNetCannyHugging FaceStable diffusionTensorText Source code Vista driving world model 2DDiffusionHuggingFaceVideo Source code Single image 3D reconstruction using MCC, SAM, and ZoeDepth 2D3DSegmentationPoint cloudSAMPaper walkthrough Source code Training a model on the LeRobot dataset 2DHuggingFaceImitation learning Source code Depth compare 2D3DHuggingFaceDepthPinhole camera Source code Mini NVS solver 2D3DHuggingFaceDepthPinhole cameraDiffusion Source code Point-E and Shap-E 3DDiffusionPointMeshPaper walkthrough Source code LLM embedding-based named entity recognition LLMEmbeddingsClassificationHugging FaceText Source code TFRecord loader 2DTensorLoaderTime series Source codeImage and video understanding Examples related to image and video processing, highlighting Rerun's 2D capabilities. View Segment anything model 2DSAMSegmentation Source codeView Detect and track objects 2DHugging faceObject detectionObject trackingOpenCV Source code Face tracking 2D3DCameraFace trackingLiveMediaPipeTime series Source code Interactive 3D annotation app with Rerun and Gradio 2D3DPinhole cameraTime seriesSAMSegmentation Source code SAM 3D body: robust Full-Body human mesh recovery 3DHuman meshBody trackingSingle-view Source code Depth compare 2D3DLidarDepthPinhole camera Source code Sam2 + DepthAnything2 2D3DHuggingFaceDepthPinhole cameraSAMSegmentation Source code PaddleOCR TextOCR2DBlueprint Source code Live camera edge detection 2DCannyLiveOpenCV Source code TAPIR: tracking any point with per-frame initialization and temporal refinement 2DPoint trackingTime seriesTensorJAXPaper walkthrough Source code DepthPro 2D3DHuggingFacePinhole cameraDepth Source code EgoExo forge 3DHuggingFaceEgocentricExocentricmanipulation Source code3D reconstruction and modelling SLAM, photogrammetry and other 3D modelling examples. View Raw mesh Mesh Source codeView Open photogrammetry format 2D3DCameraPhotogrammetry Source codeView Structure from motion 2D3DCOLMAPPinhole cameraTime series Source code KISS-ICP 3DPoint cloudLidar Source code Differentiable blocks world: qualitative 3D decomposition by rendering primitives 3DMeshPinhole cameraPaper walkthrough Source code Mast3r slam - real-time dense slam with 3D reconstruction priors 2D3DPinhole cameraTime seriesSLAM Source code Stereo vision SLAM 3DPoint cloudC++ Source code GLOMAP 3DPoint cloudGLOMAP Source code Hierarchical-Localization and GLOMAP 2D3DCOLMAPPinhole cameraTime seriesGLOMAP Source code InstantSplat 2D3DHuggingFacePinhole cameraPoint cloud Source code VistaDream: sampling multiview consistent images for single-view scene reconstruction 3DReconstructionPinhole cameraDiffusionSingle-viewGaussian splattingNovel views Source codeIntegrations Integration with 3rd party tools, formats, libraries, and APIs. View Dicom MRI TensorMRIDICOM Source code ROS node 2D3DPinhole cameraROSTime seriesURDF Source code ROS 2 bridge 2D3DPinhole cameraROSTime seriesC++ Source code ROS bridge 2D3DMeshPinhole cameraROSTime seriesC++ Source code Notebook: minimal example NotebookAPI example3D Source code Notebook: 2D neural fields NotebookNeural Field2D Source code LeRobot loader 2DVideoLoaderHugging FaceLeRobot Source code VRS viewer 2D3DVRSViewerC++ Source code Revy - Rerun integration for Bevy 2D3DGamedevBevy Source code TFRecord loader 2DTensorLoaderTime series Source code Eigen and OpenCV C++ integration 2D3DC++EigenOpenCV Source code Stock charts Time seriesBlueprint Source code ARFlow: a framework for simplifying AR experimentation workflow 3DAugmented realitySpatial computingIntegration Source code Air traffic data 2D3Dmapcrs Source code MCAP MCAPRRDROSROS 2RosbagTutorial Source code ROS TF ROSTFTransformCoordinate FrameROS 2 Source codeFeature showcase Showcase basic usage and specific features of Rerun. View IMU signals Plots Source codeView Raw mesh Mesh Source codeView Plots 2DPlotsAPI example Source codeView Graphs GraphLayoutNode-link diagramsBubble charts Source codeView Helix 3DAPI example Source code Notebook: minimal example NotebookAPI example3D Source code Notebook: viewer NotebookWidget3D Source code Notebook: viewer callbacks NotebookInteractiveCallbacks3D Source code Clock 3DAPI example Source code Compressed camera video stream 2DImage encodingVideoStreaming Source code Log file API exampleLoader Source code Lenses example Source code OpenStreetMap data MapBlueprint Source code Minimal example 3DAPI example Source code Multiprocess logging API example Source code Multithreading API example Source code Live scrolling plot PlotsLive Source code Air traffic data 2D3Dmapcrs Source code Eye control Eye control3DPinhole camera Source code Webpage WebpageJavascriptWeb-viewerTutorial Source code Any scalar Any scalarPlottingDynamicArchetype Source code GitHub Discord LinkedIn Twitter DocsBlogExamplesTeamCareers Privacy policy --- Rerun / What is Rerun? Rerun is a data platform for Physical AI that helps you understand and improve complex processes involving rich multimodal data like 2D, 3D, text, time series, and tensors.It combines simple and flexible data logging with a powerful visualizer and query engine, designed specifically for domains like robotics, spatial computing, embodied AI, computer vision, simulation, and any system involving sensors and signals that evolve over time.The problem the-problemBuilding intelligent physical systems requires rapid iteration on both data and models. But teams often get stuck because:Data from sensors arrives at different rates and in different formatsUnderstanding what went wrong requires visualizing multimodal data (images, point clouds, sensor readings) together in timeExtracting, cleaning, and preparing data for training involves too many manual stepsSwitching between different tools for each step slows everything downThe best robotics teams minimize their time from new data to training. Rerun gives you the unified infrastructure to make that happen.The Rerun Data Platform the-rerun-data-platformRerun provides an integrated solution for working with multimodal temporal data:Time-aware data model: At its core is an Entity Component System (ECS) designed for robotics and XR applications. This model understands both spatial relationships and temporal evolution, making it natural to work with sensor data, transforms, and time-series information.Built-in visualization: A fast, embeddable visualizer lets you see your data as 3D scenes, images, plots, and text—all synchronized and explorable through time. Build layouts and customize visualizations interactively or programmatically.Query and export: Extract clean dataframes for analysis in Pandas, Polars, or DuckDB. Use recordings to create datasets for training and evaluating your models.Flexible ingestion: Load data from your code via the SDK, from storage formats like MCAP, or from proprietary log formats. Extend Rerun when you need custom types or visualizations.Who is Rerun for? who-is-rerun-forRerun is built for teams developing intelligent physical systems:Robotics engineers debugging perception, controls, and planningPerception teams analyzing sensor data and model outputsML engineers preparing datasets and understanding model behaviorAutonomy teams developing and testing decision-making systemsIf you're working with robots, drones, autonomous vehicles, spatial AI, or any system with sensors that evolve over time, Rerun helps you move faster.What Rerun is not what-rerun-is-notTo set clear expectations:Not a training platform: Use Rerun with PyTorch, TensorFlow, JAX, etc. We prepare your data; you train your models.Not a deployment tool: Rerun helps you develop and understand your systems, not deploy them to production.Not a robot operating system: Rerun works with ROS, ROS2, or any other robotics stack.Not a general visualization tool: We're specialized for physical, multimodal, time-series data.How do you use it? how-do-you-use-it Use the Rerun SDK to log multimodal data from your code or load it from storageView live or recorded data in the standalone viewer or embedded in your appBuild layouts and customize visualizations interactively in the UI or through the SDKQuery recordings to get clean dataframes into tools like Pandas, Polars, or DuckDBExtend Rerun when you need toWe also offer a commercial data platform for teams that need collaborative dataset management, version control, and cloud storage. Learn more.Get started get-startedReady to speed up your iteration cycle?Quick start guide - Get up and running in minutesExamples - See Rerun in action with real dataConcepts - Learn how Rerun works under the hoodCan't find what you're looking for? cant-find-what-youre-looking-forJoin us in the Rerun Community DiscordSubmit an issue in the Rerun GitHub project GitHub Discord LinkedIn Twitter Privacy policy On this page The problemThe Rerun Data PlatformWho is Rerun for?What Rerun is notHow do you use it?Get startedCan't find what you're looking for? Edit page