Body Measurement Extraction via Deep-Learning Annotation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for extracting body measurements from images require specific poses, distances, backgrounds, and clothing, limiting their usability and accuracy.
Innovation Solution
A computer-implemented method using deep-learning networks for body segmentation and annotation, capable of extracting body measurements from 2D photos taken with a mobile device camera, regardless of pose, background, distance, or clothing type, without the need for 3D reconstruction or specialized hardware.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If 3D cameras with depth sensing are used to build 3D body models, then measurement accuracy is improved, but device complexity and accessibility deteriorate
Solution Approach 1:
The patent uses 2D camera images as copies or representations of the 3D body, applying computational algorithms to reconstruct 3D body measurements from 2D image data. This avoids the need for complex 3D depth-sensing hardware while achieving measurement accuracy through software-based 2D-to-3D reconstruction techniques.
2Adaptability or versatility
If 2D-to-3D reconstruction techniques are used with 2D videos, then body sizing can be captured, but measurement accuracy deteriorates due to template matching limitations
Solution Approach 1:
The patent segments the body into multiple key anatomical landmarks and regions (head, torso, arms, legs) and processes each segment independently to extract precise measurements. This segmented approach improves accuracy over global template matching by focusing on specific body parts with dedicated measurement algorithms.
Solution Approach 2:
The system dynamically adapts the 2D-to-3D reconstruction process based on detected body pose, clothing type, and image quality. The measurement algorithms adjust their parameters and selection of reference points in real-time to optimize accuracy for each specific user scenario, rather than using fixed template matching.
3Measurement precision
If 2D photos are used instead of 2D videos for body reconstruction, then measurement accuracy is slightly improved, but user friction and operational complexity increase
Solution Approach 1:
The patent requires only a partial view of the body in the photo (key landmarks such as head, shoulders, and torso) rather than complete 360-degree coverage. This partial action approach maintains measurement accuracy for essential body dimensions while significantly improving ease of operation, allowing users to take photos in casual poses without strict positioning requirements.
4Measurement precision
If controlled environment requirements are imposed (specific poses, distances, backgrounds, clothing), then measurement accuracy is improved, but ease of operation and adaptability deteriorate
Solution Approach 1:
The system automatically detects and adjusts for variations in拍摄 parameters including camera distance, angle, lighting conditions, and background. The algorithm changes its processing parameters dynamically based on the detected environment, allowing accurate measurements across diverse real-world conditions without requiring controlled settings.
Solution Approach 2:
The patent introduces computational algorithms as intermediaries between the 2D image and body measurements, which compensate for environmental variations. These algorithms act as mediators that correct for perspective distortion, lighting effects, and background interference, enabling accurate measurements without controlled environments.
Data Source
AI summary
Disclosed are systems and methods for full body measurements extraction using a mobile device camera. The method includes the steps of receiving one or more user parameters; receiving at least one image containing the human and a background; identifying one or more body features associated with the human; performing body feature annotation on the identified body features for generating an annotation line on each body feature corresponding to a body feature measurement, the body feature annotation utilizing an annotation deep-learning network that has been trained on annotation training data, the annotation training data comprising one or more images for one or more sample body features and an annotation line for each body feature; generating body feature measurements from the one or more annotated body features utilizing a sizing machine-learning module based on the annotated body features and the one or more user parameters; and generating body size measurements by aggregating the body feature measurements for each body feature.