Method and device for generating interactive action of digital human and object

• Main Authors: ZHOU XIAOWEI, YANG MINGHUI, PI HUAIJIN, BAO HUJUN, ZHU SHUAI
• Format: Patent
• Language: Chinese; English
• Published: 27.10.2023
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS

The embodiment of the invention relates to a digital human and object interaction action generation method and device, and the method comprises the steps: obtaining target condition data, which comprises the pose information of an initial state of a digital human, target object data, and the pose information of a target state of interaction between the digital human and a target object; determining N pieces of milestone data by using a first diffusion model, wherein the milestone is a point on a track of the digital person moving from an initial position to a target position; determining the posture of the digital person on any milestone by using a second diffusion model; determining a plurality of trajectory point data of the digital person on the trajectory between the adjacent milestones by using a third diffusion model; and determining the posture of the digital person on any track point by using a fourth diffusion model. And then determining an action sequence of the digital person moving from the initia