We are thrilled to launch our 3D point cloud labeling interface officially!
Some of you have already been using the beta version, and your feedback has been invaluable. This launch introduces many new and improved features that make 3D labeling even easier.
Enjoy the same intuitive design you know and love.
Label point clouds of any size in one place
Beyond cuboid labeling, we also support 3D keypoints, polygons & polylines, and all varieties of 3D segmentation, including semantic, instance, and panoptic segmentation.
For the vector interfaces, we support point clouds of unlimited size. The UX stays fast and responsive thanks to our backend technology. It tiles your point clouds and streams the currently visible parts to the browser, much like how Google Maps works.
Quickly label dynamic objects with Batch Mode
Labeling moving objects in a 3D scene can be a slow and frustrating experience. We support automatic interpolations, but even then a lot of frame switching is needed to make minor cuboid corrections.
Our new batch mode makes this process much smoother: it shows all occurrences of an object throughout the sequence in a single, scrollable view. Update the cuboid in one row, and see immediately how the updated interpolation adapts the cuboid positions, dimensions, and rotations in the other frames.
Accurately label static objects in the Merged Point Cloud Mode
Where batch mode helps you label moving objects, the new merged point cloud mode is unmatched when labeling static objects.
By toggling the merging mode, all frames across the sequence merge into a single big point cloud. This is particularly helpful when labeling static objects consisting of only a few points in each frame, making it hard to spot their actual dimensions. By aggregating the points across frames, the object’s outlines become clearly visible, and the object can be annotated accurately with a tight-fitting cuboid in one fell swoop.
Overlay 3D labels on top of images with sensor fusion
Do you have both point cloud and camera data available? Use them to make your labeling job easier!
You can upload data from multiple camera feeds, providing helpful context to the labelers. If you provide the camera calibration parameters, you can even overlay the point cloud and 3D labels on top of the images.
