Capabilities
- Multi-sensor MCAP/ROS playback with 8 panel types — Image, 3D / Point Cloud, Plot, Raw Messages, Log, Map, Gauge, and State Transitions. The viewer auto-detects topics from your recording and assigns each to the appropriate panel type.
- GPU-accelerated 3D point cloud rendering with 6 visualization modes — Neutral, Intensity, Rainbow, Label, Panoptic, and Image Projection. Switch between modes to inspect raw intensity values, semantic labels, panoptic instance segmentation, or LiDAR-to-camera color projection.
- Gaussian Splat scene viewer with WebGPU rendering — Load 3D Gaussian Splat reconstructions into a dedicated viewer with scene hierarchy panel, properties inspector, tool bar, mode toggle, undo/redo, and real-time statistics. GPU-based radix sorting and render bundle caching keep frame rates smooth on large scenes.
- Multi-camera synchronized playback with LiDAR-to-camera projection — View multiple camera streams in sync. Point cloud data projects onto camera images using calibrated transforms, with support for both pinhole (fx/fy/cx/cy + distortion k1-k4/p1-p2) and double-sphere (xi/alpha) camera models.
- Configurable multi-window layouts with drag-and-drop panels — The default layout arranges a topics sidebar (left), dynamic content panels (center), and a file info panel (right) in a horizontal root. Resize, rearrange, add, or remove panels to build the view you need.
- Timeline-based navigation with frame stepping and timestamp seeking — Scrub through recordings, step frame-by-frame, adjust playback speed, and jump to specific timestamps. All panels stay synchronized.
- Automatic topic detection and smart panel type assignment — When you open an MCAP file, the viewer reads the topic schema and message types, then creates the right panel for each: Image panels for camera topics, 3D panels for point clouds, Map panels for GPS, and so on.
- Pinhole and double-sphere (fisheye) camera model support — Pinhole projection (fx/fy/cx/cy) and double-sphere projection (xi/alpha) for LiDAR-to-camera overlay. Distortion parameters are stored; radial/tangential distortion correction for pinhole cameras is planned for a future update.
Supported Message Types
The viewer handles the following ROS and Foxglove message types:| Category | Messages |
|---|---|
| Camera | sensor_msgs/Image, sensor_msgs/CompressedImage, foxglove.CompressedImage, foxglove.RawImage |
| LiDAR | sensor_msgs/PointCloud2, foxglove.PointCloud |
| Radar / Laser | sensor_msgs/LaserScan, radar_msgs/RadarScan (routed via topic name keywords) |
| Transforms | tf2_msgs/TFMessage, foxglove.FrameTransform |
| Position | sensor_msgs/NavSatFix |
| IMU | sensor_msgs/Imu |
Avala reads MCAP files natively. ROS 1 bags and ROS 2 bags should be converted to MCAP format before upload. See the MCAP & ROS integration guide for conversion instructions.
If You Need…
| If you need… | Avala provides… |
|---|---|
| Play back MCAP recordings | Multi-sensor viewer with 8 panel types and synchronized timelines |
| Visualize point clouds | 6 rendering modes (Neutral, Intensity, Rainbow, Label, Panoptic, Image Projection) with WebGPU acceleration |
| View 3D scene reconstructions | Gaussian Splat viewer with scene hierarchy, properties inspector, and GPU-based rendering |
| Inspect raw sensor data | Raw Messages, Plot, Log, Gauge, and State Transition panels for every topic in your recording |
| Project LiDAR onto cameras | Automatic projection using pinhole and double-sphere camera calibrations |
| Then annotate the same data | Full annotation toolkit (bounding boxes, polygons, 3D cuboids, segmentation, polylines, keypoints, classification) without switching tools or re-uploading |
How Avala Compares
| Capability | Avala | Foxglove | Rerun |
|---|---|---|---|
| Browser-based MCAP playback | |||
| GPU-accelerated point clouds | WebGPU | WebGL | WebGPU |
| Gaussian Splat viewer | |||
| Multi-camera sync + projection | |||
| Annotation tools on same data | Full suite | ||
| Quality control workflows | |||
| Managed labeling services | |||
| Python & TypeScript SDKs | |||
| MCP server for AI assistants |
Viewers
Multi-Sensor Viewer
MCAP/ROS playback with 8 panel types, automatic topic detection, and configurable layouts for camera, LiDAR, radar, and IMU data.
3D Point Cloud Viewer
GPU-accelerated point cloud rendering with 6 visualization modes, frustum culling, LOD, and LiDAR-to-camera projection.
Gaussian Splat Viewer
WebGPU-rendered 3D Gaussian Splat scenes with scene hierarchy, properties panel, and real-time statistics.
Next Steps
MCAP & ROS
Supported message types, topic schemas, and how to prepare your recordings.
Panel Reference
Detailed documentation for each panel type: Image, Point Cloud, Plot, Map, Log, and Raw Messages.
Rendering Modes
Deep dive into the 6 point cloud visualization modes and when to use each one.
Annotation Tools
After exploring your data, start annotating with 7 professional labeling tools.