Human Key-Point Association Using BCF Feature Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing key-point association techniques, such as those described in NPL1, require defining adjacent body parts in advance, leading to potential errors and failures in associating key-points across multiple individuals in an image.
Innovation Solution
A novel key-point associating apparatus and method that utilizes Body Crosscutting Field (BCF) feature maps to connect basis and target key-points within the same individual, allowing for more accurate association by generating spatial relationships between predefined body parts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If adjacent body parts are defined in advance for key-point association, then the association process becomes simpler, but errors and failures occur when associating key-points across multiple individuals
Solution Approach 1:
The patent segments the key-point association process into two distinct stages: (1) generating individual feature maps for each target part that encode spatial relationships to basis parts, and (2) associating key-points by integrating these feature maps. This segmentation allows each stage to specialize, improving both ease of operation and reliability.
Solution Approach 2:
The patent introduces feature maps as an intermediary representation between the target image and the key-point association result. These feature maps encode spatial relationships and serve as a mediator that guides accurate association across multiple individuals, resolving the contradiction between simplicity and accuracy.
2Productivity
If existing key-point association methods are used, then processing is faster, but errors propagate when individual key-point associations are incorrect
Solution Approach 1:
The patent implements a feedback mechanism where feature maps are generated from the target image and used to guide key-point association. The association result can be fed back to refine feature map generation, creating an iterative process that improves accuracy without significantly sacrificing processing speed.
Solution Approach 2:
The patent performs preliminary generation of feature maps that encode spatial relationships before the actual key-point association. This preliminary action prepares the data structure in advance, making the subsequent association process both faster and more accurate by having pre-computed spatial guidance available.
3Use of energy by moving object
If simple key-point association is performed without feature maps, then computation is less intensive, but spatial relationships between body parts cannot be accurately captured
Solution Approach 1:
The patent applies local quality by generating separate feature maps for each target part, where each feature map encodes spatial relationships specific to that body part. This localized approach captures precise spatial relationships without requiring excessive global computation, balancing energy use and accuracy.
Solution Approach 2:
The patent transforms the key-point association problem from direct coordinate matching into a feature space representation. By projecting spatial relationships into feature maps with encoded positional information, the system captures complex spatial relationships in an additional dimensional space that is more efficient to compute than brute-force methods.
Data Source
AI summary
A key-point associating apparatus acquires a target image in which one or more persons are captured and detects, for each person, a basis key-point and one or more target key-points from the target image. The basis key-point of the person indicates a location of a basis part of the person. The target key-point of the person indicates a location of a target part of the person. The key-point associating apparatus generates a feature map for each target part based on the target image. The feature map of the target part indicates a region connecting the basis part and the target part that belongs to the person same as the basis part. The key-point associating apparatus associates, based on the feature map, the basis key-point with the target key-point that belongs to the person same as the basis key-point.