Whole-Image Identity Classification for Privacy-Preserving Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing facial recognition systems violate privacy by processing images without authorization, leading to incorrect identifications and potential violations of biometric regulations, and are inefficient when dealing with multiple individuals in a single image.
Innovation Solution
A system that processes entire images using whole-image embedding representations (WIER) trained on general image understanding tasks, preserving privacy by not isolating individuals and combining image and text embeddings to identify multiple people in a single image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If facial recognition systems process images by isolating and analyzing individual faces, then identification accuracy is improved, but privacy is violated and computational complexity increases
Solution Approach 1:
Instead of the conventional approach of isolating faces from images for recognition, this patent inverts the process by embedding entire images directly into a unified vector space. The system maps complete images preserving spatial relationships without extracting or isolating facial regions, thereby maintaining privacy while enabling recognition through whole-image comparisons in the embedded space.
2Measurement precision
If facial recognition systems process each face individually in an image, then identification precision is improved, but device complexity and processing time increase
Solution Approach 1:
The patent merges the entire image into a single embedded representation rather than processing multiple faces separately. By combining all visual information from the complete image into one unified vector in the embedded space, the system simplifies processing architecture and reduces computational complexity while maintaining the ability to perform recognition through direct vector comparisons.
3Speed
If facial recognition systems use traditional face isolation methods, then recognition speed is improved for single subjects, but productivity decreases when dealing with multiple individuals
Solution Approach 1:
The patent creates a universal processing framework where the same whole-image embedding mechanism handles both single-subject and multi-subject scenarios uniformly. The embedded space accommodates complete images regardless of the number of individuals present, allowing the system to maintain consistent processing speed and efficiency across different image types without requiring separate processing pipelines.
Data Source
AI summary
The system receives a visual representation of a scene, and without isolating an individual in the visual representation, provides the visual representation to an image feature extraction component. The system obtains from the image feature extraction component an image embedding vector representing the visual representation without isolating a single individual. The system obtains from a database a second whole-image embedding representation associated with a unique user identifier representing a user. The system determines whether the first whole-image embedding representation matches the second whole-image embedding representation. Upon determining that the first whole-image embedding representation matches the second whole-image embedding representation, the system generates an indication that the user is included in the visual representation.