Abstract:

This tutorial explains how to use Rerun's "blueprint" system to programmatically control and customize data visualization layouts. It introduces the core concept that the final Rerun viewer experience is a combination of "data + blueprint." The video demonstrates how to define layouts using the Python SDK, focusing on the central role of the application_id for persisting and reusing configurations across different sessions. The presenter walks through creating multi-panel displays with various containers (Grid, Horizontal, Vertical, Tab) and view types, including 2D/3D spatial scenes, geospatial maps, time series plots, and text logs. The tutorial covers both programmatic creation and manual adjustment of layouts directly within the viewer, highlighting how blueprints can be saved, imported, and used to create powerful, reusable templates for specific debugging scenarios in robotics and computer vision.

Controlling Visualization Layouts with Rerun Blueprints

• 0:08 What are Blueprints?: Blueprints are configurations that define the layout and properties of the Rerun viewer. They control panel visibility, the arrangement of different views (e.g., 2D/3D scenes, charts), and view-specific settings like zoom and color. The core idea is data + blueprint = viewer.
• 2:00 The application_id System: The application_id is a crucial identifier that links a specific blueprint layout to an application. All sessions sharing the same ID will automatically use the most recent blueprint associated with it, enabling persistent and consistent layouts.
• 3:12 Programmatic Control: Users can define and send blueprints directly from their code (e.g., Python using rr.send_blueprint()). This allows for creating automated, templated layouts, such as a view that shows only a specific text log and collapses all other UI panels by default.
• 5:07 Customizing View Properties: Blueprints allow for detailed configuration of individual views. The video demonstrates setting properties like a view's background color and initial zoom level (visual_bounds) programmatically.
• 8:45 Composing Complex Layouts with Containers: The system supports various containers (Grid, Horizontal, Vertical, Tab) to arrange multiple views. This enables the creation of sophisticated, information-dense dashboards combining different data types like a 2D scene, a map, and text logs in a single interface.
• 11:23 Manual Customization: In addition to programmatic control, layouts can be manually adjusted directly in the Rerun viewer. Users can drag and drop views, split panels horizontally or vertically, and create tabs to refine the layout on the fly.
• 9:57 & 13:40 Saving and Importing Layouts: Both programmatically generated and manually adjusted blueprints can be saved to a file and imported into other sessions. This feature allows teams to create, share, and reuse standardized visualization templates.
• Key Takeaway for Robotics & CV: Blueprints are powerful for creating context-specific debugging views. A developer can create a specific layout for a "perception failure" scenario and programmatically trigger it to display all relevant sensor data, point clouds, and logs when that issue occurs.