Autonomous vehicles
Autonomous vehicles

Bounding Box labeling

Bounding Box labeling

An important image annotation technique which outlines the object in the image with a box, for object classification and localization models.

  • Traffic Light Classification: Locating all traffic signals/lights in a given image by box labelling and highlighting different attributes according to various international standards
  • Sign Recognition: Labelling all traffic and other signs in a given image and highlighting their attributes that are instrumental for navigation and control functions in autonomous driving
  • Pedestrian/Animals Classification: Labelling pedestrians and animals in a given image and highlighting their static or dynamic attributes.
  • Stereo Object Detection: Labelling all objects in a given image with the objective of classifying them as static/dynamic

Semantic Annotation

Semantic Annotation

Semantic segmentation is a pixel-level labeling which identifies all the pixels in an image and segments it into its component objects for a more meaningful representation.

  • Normal Semantic Labelling: Involves pixel-level annotation of all objects in an image across 30+ classes.
  • SVS Annotation: Surround Vision Annotation which involves labelling for the surround vision camera or for fish eye image across 40+ classes.)
Autonomous vehicles
Autonomous vehicles

Skeletal/Joint Point Labeling

Skeletal/Joint Point Labeling

Identifying individual points like facial features and joint positions in the human body

  • Vulnerable Road Users: Joint Point/skeletal labelling for all vulnerable road users in a given scenario
  • Highlighting 17 joint points with attributes in a given image

Other Annotation

Other Annotation

  • Lane Detection: Marking lanes to help identify clearly demarcated lane settings and to train computer models on vehicle perception to detect a lane
  • Headlight Assistance: Marking headlights and tail lights of all visible automobiles to help detect approaching vehicles or vehicles travelling ahead
  • Lidar Labeling: Labelling output from Lidar cameras through bounding box labelling and 3D labelling
  • Other worked cases include vehicle classification, EBA (emergency break assistance) labelling etc.
  • Labeling for green staging: Bounting box labeling for green staged images which can be applied in multiple scenarios.
Automation