Rerun

What is Rerun?

Rerun is a comprehensive multimodal data stack designed to model, ingest, store, query, and visualize complex robotics-style data streams. It supports various data types including 2D, 3D, text, time series, and tensors, enabling users to debug, analyze, and improve systems involving spatial and embodied AI, robotics, simulation, and more. The platform offers SDKs in Python, Rust, and C++ for easy data logging and querying, alongside a standalone viewer for interactive visualization. Its time-aware Entity Component System data model ensures flexible yet performant handling of temporal and spatial data, making it ideal for understanding complex sensor data and system behaviors.

Key Features

• Multimodal Data Logging

  Supports logging of diverse data types such as 3D points, images, tensors, and time series with minimal code using SDKs in Python, Rust, and C++.

• Interactive Visualization

  Provides a standalone viewer and web-based options to visualize spatial and temporal data streams for debugging and analysis.

• Time-Aware Data Model

  Uses a time-aware Entity Component System (ECS) to efficiently model and query temporal and spatial relationships in data.

• Flexible Operating Modes

  Supports multiple modes including spawning a viewer, connecting to remote viewers, serving data via gRPC, and saving logs to disk.

• Open Source and Extensible

  Open source with an active community, allowing customization and extension to fit specific needs in robotics and spatial data workflows.

• Data Query and Extraction

  Enables querying of recorded data into clean dataframes compatible with tools like Pandas and DuckDB for further analysis.

Use Cases

• Robotics Debugging : Visualize and analyze sensor data streams such as lidar, camera feeds, and 3D maps to identify and resolve issues in robotic systems.
• Spatial AI Development : Track and interpret spatial representations and predictions over time to improve embodied AI models.
• Simulation and Testing : Record and review multimodal simulation data to validate and refine algorithms and system behaviors.
• Industrial Process Monitoring : Ingest and visualize complex sensor data from industrial environments for real-time monitoring and diagnostics.
• Dataset Creation : Extract clean, time-aligned datasets from logged multimodal data for training and evaluating machine learning models.

FAQs