Poselet-Based User Identification in Blurry Media
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional facial recognition techniques are inefficient when faces in media content are not clearly captured, such as when users turn their heads or are blurry, leading to inaccurate user identification.
Innovation Solution
The use of poselets and neural networks to identify users in media content by detecting poselets associated with users in images and generating multi-dimensional vectors, which are then processed to determine distance metrics and modify the neural network to improve user identification accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional facial recognition techniques are used, then user identification can be performed when faces are clearly captured, but identification accuracy deteriorates when faces are not clearly captured or are blurry
Solution Approach 1:
The patent segments the facial recognition task into multiple components: detecting poselets (key pose elements) in the image, generating multi-dimensional vectors from these poselets, and then performing recognition based on vector comparison. This segmentation allows the system to focus on distinctive pose features rather than requiring complete facial clarity, thereby maintaining accuracy even when faces are blurry or not directly facing the camera.
Solution Approach 2:
The patent transforms the recognition problem from two-dimensional image space to multi-dimensional vector space. By converting detected poselets into multi-dimensional vectors and comparing these vectors, the system can identify users based on pose characteristics rather than relying solely on visual facial clarity. This dimensional transformation enables accurate identification under conditions where traditional facial recognition fails.
2Productivity
If facial recognition is attempted with unclear or blurry faces, then more processing may be performed, but identification accuracy deteriorates
Solution Approach 1:
The patent extracts only the essential pose features (poselets) from the image rather than processing the entire image data. By detecting and extracting these key pose elements and converting them to vectors, the system performs targeted processing that is both efficient and accurate, avoiding the need to process all image data while maintaining identification precision even with blurry faces.
Data Source
AI summary
Systems, methods, and non-transitory computer-readable media can receive a first image including a representation of a first user. A second image including a representation of a second user can be received. A first set of poselets associated with the first user can be detected in the first image. A second set of poselets associated with the second user can be detected in the second image. The first image including the first set of poselets can be inputted into a first instance of a neural network to generate a first multi-dimensional vector. The second image including the second set of poselets can be inputted into a second instance of the neural network to generate a second multi-dimensional vector. A first distance metric between the first multi-dimensional vector and the second multi-dimensional vector can be determined.