Key-Point Association Using Direction Maps for Faster Grouping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing key-point association techniques, such as those using Part Affinity Fields (PAF), often include regions apart from the key-points, leading to slow convergence during training and inefficiencies in key-point grouping.
Innovation Solution
A key-point associating apparatus and method that generates spatial feature maps with separate direction regions for each key-point pair, avoiding regions between key-points and enabling efficient grouping of key-points belonging to the same person.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If PAF includes regions apart from key-points (such as middle point regions), then the feature map can represent directional information between key-points, but training convergence becomes slow
Solution Approach 1:
The patent extracts only the necessary directional information from key-point regions and removes redundant middle point regions from the feature map. By generating direction maps that contain only directional vectors at key-point locations rather than filling entire regions between key-points, the solution eliminates unnecessary computational data while preserving essential directional relationships for key-point association.
2Loss of information
If PAF fills regions between key-points with directional pixel values, then directional relationships are represented, but computational efficiency decreases
Solution Approach 1:
The patent segments the feature representation into discrete key-point locations rather than continuous regions. Each key-point has an associated direction map containing only the directional vector to its paired key-point, rather than filling the entire spatial region between them. This segmentation reduces the total number of pixels processed while maintaining directional relationship information.
3Loss of information
If the feature map includes redundant regions apart from key-points, then directional information is preserved, but key-point grouping accuracy may be compromised
Solution Approach 1:
The patent extracts precise directional vectors at key-point locations and removes redundant middle point regions that could introduce noise or ambiguity. By concentrating directional information only at detected key-point positions rather than distributing it across intermediate regions, the solution improves the precision of key-point pairing decisions while maintaining complete directional information for association.
Data Source
AI summary
A key-point associating apparatus acquires a target image on which one or more persons are captured, detects key-points from the target image, and generates a spatial feature map for each one of pairs of the body parts. The spatial feature map includes a first direction region for each key-point that represents a first body part of the corresponding pair and the second direction region for each key-points that represents a second body part of the corresponding pair. The first and second direction regions belonging to a same person as each other represent a direction from the key-point of the first direction region to the key-point of the second direction region. The key-point associating apparatus generates a key-point group for each one of the persons captured on the target image.