Human Figure Recognition via Joint Score Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for recognizing human figures in images, such as those discussed by Newell et al., face challenges in correctly connecting joints of different human figures when multiple figures are adjacent to each other, leading to misidentification and incorrect posture estimation.
Innovation Solution
An information processing apparatus that uses a pre-trained model to estimate joint positions and scores, inferring scores on lines connecting joints to distinguish between different human figures, and employing a score map to identify joint groups and recognize postures by integrating joint maps with close scores.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a machine learning-based recognition technique using a pre-trained model is used to recognize objects in images, then recognition speed and efficiency are improved, but accuracy deteriorates when multiple objects are adjacent to each other due to joint misconnection
Solution Approach 1:
The patent introduces an intermediary verification process that checks the validity of joint connections after initial detection. A verification unit validates whether detected joints belong to the same object by examining spatial relationships and connection patterns, preventing misconnection errors while preserving the speed benefits of pre-trained models
Solution Approach 2:
The system implements feedback mechanisms where detection results are validated and corrected through verification processes. The verification unit provides feedback on joint connection validity, and incorrect connections are corrected by re-evaluating joint relationships based on object boundaries and spatial constraints
2Device complexity
If joints are connected based on close scores without verification, then processing complexity is reduced, but reliability deteriorates due to misidentification of adjacent objects
Solution Approach 1:
The patent performs preliminary actions by establishing object boundaries and validating joint relationships before final connection is made. The verification unit checks spatial relationships and connection validity in advance, preventing unreliable connections while maintaining processing efficiency through structured verification protocols
Data Source
AI summary
An apparatus that estimates a position of each object in image data in which a plurality of objects is imaged, the apparatus includes a first acquisition unit configured to acquire position information indicating positions of joints of the plurality of objects in the image data, a second acquisition unit configured to acquire a score map in which a feature for identifying each object is converted into a numerical value, the score map being output by a pre-trained model in response to input of the image data, and an identification unit configured to identify positions of joints belonging to each of the plurality of objects, based on the position information and the score map.