This video tutorial serves as the second part of the "Gentle Introduction to Rerun" series, focusing on the platform's capabilities for visualizing and querying time-series data. The presenter demonstrates how to create a real-time animation of rotating 2D points to explain core concepts. Key topics covered include logging multimodal data (2D points, colors, scalar values) on a frame-by-frame basis, structuring data using hierarchical entity paths, and distinguishing between static and dynamic data. The tutorial also explores the interactive Rerun viewer, showcasing how to use the timeline to play and scrub through data, and how to customize the layout with multiple synchronized views, including a 2D spatial plot, scalar charts, and a raw data frame inspector. The overall goal is to equip developers in fields like robotics and AI with the skills to understand how their data evolves over time.
Visualizing Temporal Data with Rerun: A Step-by-Step Guide
0:00 Introduction to Time in Rerun: The video introduces its core topic: visualizing data that changes over time. It builds on the static data visualization from Part 1 by creating a dynamic animation of rotating points over 360 frames.
1:09 Live Animation Demo: A preview of the final result shows four points rotating in a 2D space. The demonstration highlights the interactive timeline at the bottom, which allows the user to play, pause, loop, and scrub through the recorded frames of the animation.
2:54 Logging Static vs. Dynamic Data: The code walkthrough begins by showing how a static origin point is logged once, outside the main loop. This is contrasted with the animated points, which are logged repeatedly inside the loop to create the motion.
3:22 Frame-by-Frame Logging: The core of the animation is a loop that runs for 360 frames. In each iteration, the script calculates new positions, colors, and sizes for the points and logs these attributes to Rerun. This action of logging data within a loop is fundamental to creating time-series visualizations in Rerun.
4:32 Logging Multimodal Data: The tutorial explains that Rerun can log more than just spatial data. In each frame, scalar values for each point's X/Y position and rotation angle are also logged. This automatically generates synchronized time-series plots that visualize how these values change over time.
6:10 Understanding Entity Paths: A key concept of Rerun is introduced: organizing data with hierarchical entity paths (e.g., position/point_0/x). This allows for structured, queryable data, enabling users to drill down from a high-level entity (like a robot) to a specific component (like a sensor's X-coordinate).
7:29 Synchronized Multi-View Layout: The video demonstrates the power of Rerun's customizable interface ("Blueprint"). It shows how multiple views—the 2D spatial view, the scalar plots, and a raw Dataframe view—can be displayed simultaneously. All views are synchronized, so scrubbing the timeline updates the visualization across all panels, providing a comprehensive understanding of the data at any given moment.
8:40 Conclusion and Next Steps: The tutorial concludes by emphasizing how these features are critical for debugging and understanding complex systems in robotics and machine learning. It also provides a preview of Part 3, which will cover coordinate transforms.