Imitating motion capture clips using a neural network

A video generator system, comprising:

• a physics simulator to adjust movement of a target object using a stability threshold and a movement agent, and to generate a frame of the target object;
• and
• a neural network (NN) to provide movement adjustments to the movement agent of the target object by indicating one or more of the movement adjustments utilizing one or more target angles of joints of a reference object from a motion capture (MOCAP) video clip
• wherein the NN is at least one of a tracking NN or a recovery NN and when the stability threshold is not satisfied, the movement agent is a recovery agent and the NN is the recovery NN.

A computer program product having a series of operating instructions stored on a non-transitory computer-readable medium that directs a data processing apparatus when executed thereby to perform operations to imitate a reference object from a motion capture (MOCAP) video clip using a physics simulator and a neural network (NN), comprising:

• tracking the reference object using a movement agent of a target object;
• and
• adjusting a movement of the target object using a stability threshold and the movement agent, wherein the movement agent utilizes an output from the NN to provide modifications to the movement:
• and
• compensating for signal noise within the MOCAP utilizing the movement agent and the NN,
• wherein the output is indicative of at least one of:
• one or more of torque parameters, one or more angle and joint positioning parameters, one or more applied force parameters, or one or more gain parameters of one or more proportional derivative (PD) controllers.