Sama delivers point cloud annotations in JSON. The JSON contains an array of tasks, in which each element of the array is a 3D task and its annotations.

Cuboid

Cuboid annotations have two formats, both are present in the delivery file

Format 1
Points (x,y,z) coordinate data for the cuboid's eight vertices.  Contextual labels   Unique object IDs Visibility information(is the object fully visible, occluded, or not visible in a particular frame)
Format 2
Position_center (x,y,z)   Dimensions (length, width, height)   Direction (roll, pitch, yaw)  Contextual labels   Unique object IDs   Visibility information(is the object fully visible, occluded, or not visible in a particular frame)

Polylines

Format
An array of ordered pairs (x,y) coordinate data for each line vertices.  Contextual labels   Unique object IDs Visibility information(is the object fully visible, occluded, or not visible in a particular frame)

Fused Annotation

Format
One array per camera (left camera, right_camera, center_camera)  Contextual labels   Unique object IDs Visibility information(is the object fully visible, occluded, or not visible in a particular frame)