Preparing data for high-precision absolute localization of a moving object along a trajectory

A method comprising:

• selecting a first frame in first video footage;
• selecting a second frame in second video footage;
• generating a first velocity, along a first dimension, in the direction of yaw of a first moving object that is associated with the first frame;
• generating a second velocity, along a second dimension, in the direction of yaw of the first moving object;
• generating a first particular velocity, along the first dimension, in the direction of yaw of a second moving object that is associated with the second frame;
• generating a second particular velocity, along the second dimension, in the direction of yaw of the second moving object;
• generating a ratio based on the first velocity, the second velocity, the first particular velocity, and the second particular velocity;
• generating a delta yaw value based on an inverse tangent of the ratio;
• including the delta yaw value in a training instance for training a machine-learned model.

One or more non-transitory storage media storing instructions which, when executed by one or more computing devices, cause:

• identifying a sliding window of a set of adjacent points along a trajectory of a moving object;
• identifying a midpoint of the set of adjacent points;
• based on the set of adjacent points, generating a first polynomial equation for a first dimension, wherein a first value of the first polynomial equation at a particular timestamp associated with the midpoint is a first newly calculated position value along the first dimension and a first derivative at the particular timestamp is a first velocity along the first dimension;
• based on the set of adjacent points, generating a second polynomial equation for a second dimension that is different than the first dimension, wherein a second value of the second polynomial equation at the particular timestamp is a second newly calculated position value along the second dimension and a particular first derivative at the particular timestamp is a second velocity along the second dimension;
• generating a velocity in direction of yaw based on the first velocity, the second velocity, and a slip angle associated with the midpoint;
• generating a yaw angle based on the velocity in direction of yaw;
• including the yaw angle in a training instance that includes an image representing a pose of the moving object.