CPC G06T 17/20 (2013.01) [G06N 3/045 (2023.01); G06N 3/08 (2013.01); G06T 7/20 (2013.01); G06T 7/70 (2017.01); H04N 5/04 (2013.01); G06T 2207/30196 (2013.01); G06T 2207/30244 (2013.01)] | 20 Claims |
1. A method of predicting mesh data, comprising:
receiving first two-dimensional (2D) image data representing a human body, wherein the first 2D image data is captured by a first image sensor having a first field-of-view;
receiving second 2D image data representing the human body, wherein the second 2D image data is captured by a second image sensor having a second field-of-view different from the first field-of-view;
determining, by inputting the first 2D image data into a first three-dimensional (3D) mesh prediction model, a first pose parameter θ1 and a first shape parameter β1;
determining, by inputting the second 2D image data into a second 3D mesh prediction model, a second pose parameter θ2 and a second shape parameter β2;
determining a first difference value representing a difference between the first shape parameter β1 and the second shape parameter β1;
determining a second difference value representing a difference between the first pose parameter θ1 and the second pose parameter θ2;
updating the first 3D mesh prediction model using the first difference value and the second difference value to generate an updated 3D mesh prediction model; and
generating, by the updated 3D mesh prediction model, a 3D representation of the human body.
|