Virtual Avatar Generation Using Deep Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for generating virtual avatars in live broadcasts and short videos require a long time to map facial features and expressions to animated images, as they typically use fixed animated images, limiting efficiency and accuracy.
Innovation Solution
A method and apparatus that acquire an initial avatar, determine its expression parameters, and use pre-trained deep neural networks, specifically generative adversarial networks, to generate a target virtual avatar associated with the avatar's attribute and expression, including parameters for five sense organs, such as eyes and mouth, to efficiently produce a virtual avatar.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If fixed animated images are used to map facial features and expressions, then the mapping can be achieved, but the time required for generating an animation similar to a real person becomes long
Solution Approach 1:
The patent changes the parameter representation from fixed animated images to expression parameters that describe facial features and expressions. By using expression parameters (such as eye opening degree, mouth shape, head angle) instead of fixed images, the system can efficiently generate avatars that accurately reflect the user's facial characteristics without requiring long processing times.
Solution Approach 2:
The patent creates a virtual avatar that copies the user's facial features and expressions by extracting key point coordinates and expression parameters from the user's face image. The virtual avatar is generated by applying these copied parameters to a three-dimensional avatar model, enabling fast and accurate replication of facial characteristics.
2Measurement precision
If expression parameters of five sense organs are determined, then the accuracy of virtual avatar generation is improved, but the complexity of processing increases
Solution Approach 1:
The patent segments the facial expression analysis into separate components for different sense organs (eyes, nose, mouth, ears, head). Each organ's expression parameters are detected independently by identifying key point coordinates on the face image, which simplifies the overall processing complexity while maintaining comprehensive accuracy.
Solution Approach 2:
The patent introduces key point coordinates as an intermediary representation between the raw face image and the final virtual avatar. By first detecting key point coordinates for each sense organ and then using these coordinates to determine expression parameters, the system simplifies the processing complexity while maintaining high accuracy.
Data Source
AI summary
A method and apparatus for generating a virtual avatar are provided. The method may include: acquiring a first avatar, and determining an expression parameter of the first avatar, where the expression parameter of the first avatar including an expression parameter of at least one of five sense organs; and determining, based on the expression parameter of at least one of the five sense organs, a target virtual avatar that is associated with an attribute of the first avatar and has an expression of the first avatar.