GAN-Based Face Image Augmentation for Pose and Attribute Diversity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Obtaining multiple poses and attribute variations of face images for a face database is costly and inefficient in existing methods.
Innovation Solution
A method using a Generative Adversarial Network (GAN) to generate additional face images by determining face direction, transforming into feature vectors, applying facial attribute and pose conversion algorithms, and filtering using classifiers to ensure quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If multiple poses and attribute variations of face images are obtained through additional costly data capture, then the quantity and diversity of face images in the database is improved, but the cost and time consumption increase
Solution Approach 1:
The patent uses a generative adversarial network to create synthetic copies of face images with varied poses and attributes. The GAN generates realistic face images that mimic the characteristics of real captured images, eliminating the need for costly additional data capture while increasing the quantity and diversity of the face database
Solution Approach 2:
The patent transforms face images into feature vectors and modifies specific parameters such as pose angles, lighting conditions, and facial attributes through algorithmic processing. By changing these parameters in the feature space and reconstructing images, the system generates diverse face variations without physical re-capture
2Quantity of substance
If multiple poses and attribute variations of face images are obtained through additional costly data capture, then the quality and diversity of face images in the database is improved, but the time consumption increases
Solution Approach 1:
The generative adversarial network rapidly generates synthetic face image copies through computational processes, producing multiple pose and attribute variations in minutes rather than the hours or days required for physical re-capture sessions
Solution Approach 2:
The system performs preliminary transformation of face images into feature vectors and establishes the generative model in advance. Once trained, the model can quickly generate diverse face variations on demand without requiring time-consuming additional data capture sessions
3Adaptability or versatility
If facial attribute editing and pose conversion algorithms are applied to generate varied face images, then the diversity of face images is improved, but the system complexity increases
Solution Approach 1:
The patent introduces feature vectors as an intermediary representation between real face images and generated variations. By transforming images into compressed feature vectors and performing edits in this intermediate space, the system achieves complex attribute modifications and pose conversions through manageable algorithmic operations
4Manufacturing precision
If classifiers and recognizers are used to filter generated face images, then the quality and accuracy of the face database is improved, but the processing time increases
Solution Approach 1:
The patent applies a multi-stage filtering process where classifiers and recognizers perform partial verification at different stages. Rather than exhaustive validation of every generated image, the system uses successive filters to quickly eliminate low-quality results while preserving potentially useful variations for further processing
Data Source
AI summary
The present invention provides a method for increasing face images. The method includes: obtaining a face image, and determining whether the face image belongs to a frontal direction or a side direction; transforming the face image into a feature vector; performing a facial attribute editing algorithm according to the feature vector to generate a first generation vector which is different from the face image in at least one attribute; if the face image belongs to the side direction, performing a facial pose conversion algorithm according to the feature vector to generate a second generation vector which is different from the face image in pose directions; inputting the first generation vector or the second generation vector to a generative adversarial network to output at least one generated face image; and adding the face image and the generated face image into a database.