KnowledgeApr 2, 2026·3 min read

Rerun — Visualize Multimodal AI Data in Real-Time

SDK for logging, storing, and visualizing 3D, images, time series, and text in real-time. Built for robotics and AI. 10K+ stars.

TO
TokRepo精选 · Community
Quick Use

Use it first, then decide how deep to go

This block should tell both the user and the agent what to copy, install, and apply first.

```bash pip install rerun-sdk ``` ```python import rerun as rr import numpy as np rr.init("my_app", spawn=True) # Opens the Rerun viewer # Log different data types rr.log("camera/image", rr.Image(np.random.rand(480, 640, 3))) rr.log("lidar/points", rr.Points3D(np.random.rand(1000, 3))) rr.log("metrics/loss", rr.Scalar(0.42)) rr.log("text/status", rr.TextLog("Model loaded successfully")) ``` The Rerun Viewer opens automatically with interactive 2D/3D visualization. ---
Intro
Rerun is an open-source SDK with 10,500+ GitHub stars for logging, storing, querying, and visualizing multimodal data streams in real-time. Built for robotics, computer vision, and AI development, it handles 3D point clouds, camera images, time series, text logs, tensors, and more — all in a single interactive viewer. Log data from Python, Rust, or C++ and Rerun automatically handles time synchronization, layout, and interactive exploration. Think of it as a modern replacement for ad-hoc matplotlib scripts and custom visualization code. Works with: Python, Rust, C++, ROS 2, any data pipeline. Best for robotics engineers, CV researchers, and AI developers who need to visualize complex multimodal data. Setup time: under 2 minutes. ---
## Rerun Data Types & Features ### Supported Data Types | Type | API | Use Case | |------|-----|----------| | **Images** | `rr.Image()` | Camera feeds, segmentation masks | | **3D Points** | `rr.Points3D()` | LiDAR, depth sensors, embeddings | | **3D Meshes** | `rr.Mesh3D()` | 3D models, reconstructions | | **Bounding Boxes** | `rr.Boxes2D/3D()` | Object detection | | **Time Series** | `rr.Scalar()` | Loss curves, metrics | | **Text** | `rr.TextLog()` | Logs, status messages | | **Tensors** | `rr.Tensor()` | Feature maps, attention weights | | **Transforms** | `rr.Transform3D()` | Coordinate frames | | **Arrows** | `rr.Arrows3D()` | Velocities, forces | ### Real-Time Streaming ```python import rerun as rr import time rr.init("live_demo", spawn=True) for i in range(1000): rr.set_time_sequence("frame", i) rr.log("sensor/temperature", rr.Scalar(20 + np.sin(i/10))) rr.log("camera/frame", rr.Image(capture_frame())) rr.log("robot/position", rr.Points3D([[x, y, z]])) time.sleep(0.033) # 30 FPS ``` ### Time Scrubbing The viewer includes a timeline scrubber — scroll through time to see how all your data changes together. Images, 3D points, metrics, and text all stay synchronized. ### Blueprint API ```python import rerun.blueprint as rrb # Define custom viewer layout blueprint = rrb.Vertical( rrb.Spatial3DView(origin="world"), rrb.Horizontal( rrb.Spatial2DView(origin="camera/image"), rrb.TimeSeriesView(origin="metrics"), ), ) rr.init("my_app", spawn=True) rr.send_blueprint(blueprint) ``` ### Use Cases | Domain | What to Visualize | |--------|------------------| | **Robotics** | Camera feeds + LiDAR + robot state + trajectory | | **Computer Vision** | Input images + detections + segmentation masks | | **SLAM** | 3D maps + camera poses + feature points | | **LLM/AI** | Attention weights + embeddings (t-SNE) + metrics | | **Simulation** | Physics bodies + sensor readings + agent actions | --- ## FAQ **Q: What is Rerun?** A: Rerun is an open-source SDK with 10,500+ GitHub stars for logging and visualizing multimodal data (3D, images, time series, text) in real-time, with an interactive viewer that syncs all data streams on a shared timeline. **Q: How is Rerun different from TensorBoard?** A: TensorBoard focuses on ML training metrics (scalars, histograms). Rerun handles multimodal data — 3D point clouds, camera images, robot state, and metrics all in one viewer with time synchronization. It's built for robotics and CV, not just training logs. **Q: Is Rerun free?** A: Yes, open-source under Apache-2.0. Rerun also offers a commercial product (Rerun Pro) with collaboration features. ---
🙏

Source & Thanks

> Created by [Rerun](https://github.com/rerun-io). Licensed under Apache-2.0. > > [rerun](https://github.com/rerun-io/rerun) — ⭐ 10,500+

Discussion

Sign in to join the discussion.
No comments yet. Be the first to share your thoughts.

Related Assets