Raw recordings in.
Metric 3D data out.
Upload a .mcap or ROS1 .bag and get back a metric 3D point cloud, reconstructed from LiDAR and GNSS, or from a single camera with monocular metric depth. Runs in your browser. No ROS install, no SLAM tuning, no Docker.
Works with any LiDAR that publishes sensor_msgs/PointCloud2.
Representative point cloud, coloured by elevation.
Show the recording.
Get the map.
The whole pitch, in engineer native language. Feed LidarFlow the same topics you already record. Download the mapped artifacts you actually need.
What you upload
- recording.mcap…or run.bag, a LiDAR recording with the topics you already capture.
- /pointcloud_topicAny sensor_msgs/PointCloud2 topic. Validated before the run starts.
- /imu/dataOptional. Used when the recording includes an IMU.
- /fixOptional GNSS topic. Needed only when you want a georeferenced map.
What you get back
- output.pcdThe mapped point cloud artifact used across the product.
- map.plyThe canonical full GLIM export, when the precision run produces it.
- metadata.jsonRun settings, validation, and provenance in one file.
- slam_trajectory.txtIncluded whenever trajectory material is available for the run.
Upload. Validate. Reconstruct. Georeference.
The deep walkthrough lives on its own page. The short version is four verbs and one browser tab.
Upload
Drop a .mcap or ROS1 .bag, up to 20 GB. No ROS install required.
Validate
We detect and check LiDAR, TF, GNSS and IMU topics, and tell you what we found before the run.
Reconstruct
GLIM SLAM builds the map from LiDAR, or monocular metric depth builds it from a single camera.
Georeference
FlexCloud aligns the result to WGS84 when GNSS is present, so the map lands in the real world.
Start with the page that matches the job you need to do.
These guides go after the exact queries robotics engineers type when they need to turn a recording into a usable map.
Rosbag to PCD
Compare the local CLI, a Python script, and the browser workflow when you need a point cloud out of a recording.
Open guideRosbag to point cloud map
See the difference between exporting frame-by-frame PCD files and building one mapped point cloud you can review.
Open guideMCAP to point cloud
Inspect MCAP with upstream tools, then bridge the gap from recorded topics to a mapped output in the browser.
Open guideLidarFlow vs DIY pipeline
Compare running GLIM + FlexCloud yourself against using a managed rosbag-to-map workflow.
Open guideQuestions engineers ask before handing over a recording.
The things we refuse to fake.
Most rosbag tooling lives or dies on trust. We would rather lose a deal than ship a number we cannot back up. Here is what that looks like in practice.
We publish numbers we can reproduce.
You will not find a marketing chart of accuracy versus a competitor on this site. Reconstruction quality depends on your sensor stack, route, and trajectory. We will quote numbers from your own recording, in your inbox, not a billboard.
ROS 2 .db3 is not supported yet.
Today the upload validator accepts ROS1 .bag and .mcap. ROS 2 .db3 is on the roadmap with a waitlist below. We would rather tell you no than silently fail on upload.
We do not train on your recordings.
Uploads stay in your workspace. We do not sample frames for model training, we do not share them with sensor vendors, and we delete them on the retention you set.
One click to cancel from the dashboard.
No retention call, no five-step downgrade flow. If the workflow stops earning its keep, you turn it off the same way you turned it on.
Drop the bag.
Get the map.
Free while we launch. Create a workspace, upload your first recording, download the map. The whole thing takes under five minutes, and there is no card to enter.