Multi-modal versatility
Annotate camera, LiDAR, and radar together, in 2D and 3D, to reconstruct the full scene around the vehicle.
Automotive
Build perception systems that read the road.
A professional platform for annotating sensor, LiDAR, and video data into precise, ML-ready training datasets.
Applications
Driver monitoring systems (DMS)
Detect driver fatigue, distraction, or unconsciousness from annotated facial and posture data, with stable tracking even in low cabin light.
Technical edge
AI-powered automation
Pre-label vehicles, pedestrians, and signs, then interpolate boxes and tracks across long driving sequences.
Quality control
Validation and multi-level review catch missed road users and tracking drift before data reaches training.
Collaborative workflow
Assign frames and sequences across the team, with roles for annotators, reviewers, and leads on fleet-scale datasets.
Customization & integration
Add custom classes and attributes for your operational design domain, and connect to your perception stack through the API.
Enterprise-grade security
Protect pre-release sensor data with strict confidentiality and on-premise deployment when required.
Annotation types
Bounding box
Rectangular boxes that locate and classify vehicles, pedestrians, and road signs in 2D.
Oriented bounding box
Rotated rectangles that capture the size and heading of angled vehicles.
Polygon
Flexible shapes that outline roadway boundaries, work zones, and irregular objects.
Points
Single points that track a driver's gaze or distant roadside objects.
Lines & multilines
Trace road markings, lane boundaries, curbs, and barriers.
Skeletons
Connected keypoints that model pedestrian or passenger poses to predict actions.
Cuboid
3D boxes that capture an object's dimensions to calculate distance in 3D space.
3D point cloud
Spatial labeling of LiDAR data to build accurate point clouds around the vehicle.
Semantic segmentation
Pixel-level classification of the whole scene into road, pavement, vegetation, and sky.
Instance segmentation
Pixel-wise separation of each object within a class.
Bitmap
Raster masks that isolate complex structures like fences or tree branches.
Mesh
Polygonal meshes that build 3D road models and digital twins of urban environments.
Custom
A mix of annotation methods built around your specific ADAS or autonomy task.
Put your automotive data to work
Book a demo and we'll scope your perception dataset with you.
Custom solution? hello@keylabs.ai
