Animation Retargeting via Canonical Space Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current animation retargeting methods are inefficient and time-consuming, particularly when translating motion capture data from an actor's face to a computer-generated character's face with different proportions and personality, as they struggle to accurately adapt facial expressions due to geometric differences.
Innovation Solution
The system creates a mapping between the source and target objects' spaces using a training set of corresponding shapes, allowing for interpolation and extrapolation of facial expressions, and provides a user interface for refining this mapping based on feedback, enabling accurate transformation of facial expressions through affine transformations and neighborhood-based mappings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional animation retargeting methods are used to translate motion capture data from an actor's face to a computer-generated character's face, then the process can be completed, but the accuracy of facial expression adaptation deteriorates due to geometric differences between faces
Solution Approach 1:
The patent transforms facial geometry into a standardized parameter space (canonical space) where geometric differences are normalized. By representing facial expressions as parameters in this canonical space rather than direct vertex coordinates, the system can accurately adapt expressions across faces with different geometries. The parameter space transformation enables precise control of facial expression adaptation while handling geometric variations.
Solution Approach 2:
The patent introduces a canonical space as an intermediary representation between the source and target facial geometries. This canonical space acts as a mediator that decouples the geometric differences from the expression transfer process, allowing expressions to be adapted accurately regardless of the underlying geometric variations between actor and character faces.
2Productivity
If traditional animation retargeting methods are used to adapt facial expressions, then the process can be completed, but the time required increases due to inefficiency in handling geometric transformations
Solution Approach 1:
The patent performs preliminary actions by pre-computing the parameter space transformation and establishing the canonical representation of facial geometry before the actual expression retargeting occurs. By preparing the parameter space mappings and canonical models in advance, the system significantly reduces the time required for real-time or near-real-time expression adaptation, improving overall retargeting efficiency.
3Ease of manufacture
If direct vertex mapping is used to transfer facial expressions, then the process is simple, but the accuracy deteriorates when source and target faces have different proportions and structures
Solution Approach 1:
Instead of directly mapping vertices between source and target faces, the patent transforms the problem into parameter space where expressions are represented as controlled parameters. This parameter-based approach maintains simplicity by using standardized representations while achieving high accuracy in expression transfer, even when faces have different proportions and structures.
Data Source
AI summary
The subject matter of this specification can be embodied in, among other things, a method that includes determining a transform of a portion of constituent components of a source shape. The transform includes one or more states for the portion of constituent components of the source shape. The method also includes accessing a mapping function that associates the one or more states with one or more controls for a target shape, where the one or more controls configured for access by a user for manipulating constituent components of the target shape. The method includes outputting a transform for the target shape based on the one or more controls associated with the transformed constituent components of the source shape.