Iterate faster on robotics learning with unified infrastructure. Ingest, visualize, annotate, query, and transform - from collection to training.
Rerun is open source
6,200
Visualize consistently as you explore, build, evaluate, and debug
Manage and share all data and visualizations in one place
Run snappy dataframe queries optimized for robotics data
Transform logs into training-ready data with simple pipelines
Ingest, store, and index robotics logs of multi-formats
Search and retrieve petabyte-scale robotics data in seconds
Powerful and flexible visualization for spatial and embodied AI that's shockingly easy to get started with.
pip install rerun-sdk
rerun
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.
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 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 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 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 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 →
Rerun accelerates your speed of progress by helping you SEE faster
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.
# Find rows around the time of the anomaly
around_anomaly = dataset.filter(
dn.Expr.between(
col("time"),
anomaly_time - 10,
anomaly_time + 10
)
)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.
Transform messy logs into training-ready datasets with a consistent and flexible pipeline.
Debug visually as you train, evaluate, and iterate ML models.
Visualize robot runs and turn them into training data locally.
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))Built on open source, extended for production. Consistently produce large-scale, high-quality training data from massive robot runs.
# 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]
)
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
Visualize robot runs and turn them into training datasets, run locally
Dual licensed under MIT and Apache 2
Join on GitHubTurn large scale robotics log data into high-quality training data
Book a demo
Book a demoVisualize robot runs and turn them into training datasets, run locally
Dual licensed under MIT and Apache 2
Join GithubTurn large scale robotics log data into high-quality training data
Beta release with selected customers
Book a demo
Robotics teams are trying to scale end-to-end learning without the mature data infrastructure that made it work for LLMs. This post is a map of where the tax comes from, walking backwards from evaluation to collection.
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 brings major performance improvements, reflection-based ROS2 MCAP support, experimental lenses, and continued polish across the viewer and SDK.
