Head Region Digital Representation Editing for Gaze Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Images or digital representations of a head region often exhibit incorrect gaze direction due to the offset positioning of capture and display devices, leading to disconcerting interactions in teleconferencing and similar systems, and existing solutions are computationally expensive or inefficient.
Innovation Solution
A method involving two machine learning algorithms is employed to generate reference data for adjusting digital representations of head regions, using a first algorithm to provide highly detailed data and a second algorithm to generate less detailed, compressed data for efficient processing, thereby improving accuracy and reducing computational demands.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a single machine learning algorithm is used to generate reference data for adjusting digital representations, then comprehensive adjustment capabilities are achieved, but computational complexity and processing time increase significantly
Solution Approach 1:
The patent divides the machine learning processing into two separate algorithms: a first machine learning algorithm that generates comprehensive reference data, and a second machine learning algorithm that processes input images using that reference data. This segmentation allows the system to maintain comprehensive adjustment capabilities while reducing the computational burden on any single processing stage, as each algorithm can be optimized for its specific function rather than requiring one algorithm to handle all adjustment tasks.
2Manufacturing precision
If highly detailed reference data is generated for accurate adjustment, then adjustment precision improves, but data storage and bandwidth requirements increase
Solution Approach 1:
The patent extracts only the essential and most relevant features from the reference data generated by the first machine learning algorithm, rather than transmitting or storing all the detailed data. The second machine learning algorithm is trained to work with this extracted subset of information, maintaining adjustment precision by focusing on the most critical parameters while significantly reducing data storage and bandwidth requirements.
3Productivity
If minimal processing is used to reduce computational demands, then processing speed improves, but adjustment accuracy deteriorates
Solution Approach 1:
The patent performs preliminary processing by generating comprehensive reference data using the first machine learning algorithm before the actual adjustment operation. This pre-computed reference data is then used by the second machine learning algorithm, which can perform rapid adjustments without needing to recalculate complex transformations in real-time. The computationally intensive work is done in advance, allowing fast processing during actual use while maintaining high accuracy.
Data Source
AI summary
A methods for adjusting a digital representation of a head region includes identifying a target patch in the digital representation of the head region where the target patch includes a target feature of the digital representation of the head region, determining, by a first inferential model and based on a feature of the target patch, a displacement field configured to modify a relative position of a visual element of the target feature within the target patch, determining, by a second inferential model and based on the displacement field and the feature of the target patch, editing instructions configured to modify the visual element, and adjusting the target patch in the digital representation of the head region based on the editing instructions.


