2-D Pose Prediction With User-Guided Keypoint Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Motion tracking systems face challenges in accurately identifying 2-D poses due to visual occlusion, overlapping body images, and poor image quality, leading to failures in keypoint detection.
Innovation Solution
A machine learning (ML) model-based system for 2-D pose prediction and correction that allows users to intuitively identify and correct keypoint detection errors, using ML model-based techniques for pose editing, labeling, and guiding the prediction process through iterative manipulation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If automated keypoint detection is used, then productivity is improved, but measurement precision deteriorates due to visual occlusion, overlapping body images, and poor image quality
Solution Approach 1:
The system implements an iterative feedback mechanism where initial automated keypoint predictions are generated, then user corrections are incorporated back into the system. The ML model reprocesses the image with the corrected keypoints as input, generating refined predictions that maintain automated efficiency while improving accuracy through user feedback loops.
Solution Approach 2:
The patent introduces an intermediary ML model that acts as a bridge between automated detection and manual correction. The model takes user-corrected keypoints as input and generates refined predictions, serving as a mediator that translates user corrections into improved automated results without requiring complete manual annotation.
2Measurement precision
If manual correction of keypoint detection errors is implemented, then measurement precision is improved, but loss of time increases due to iterative editing requirements
Solution Approach 1:
The system applies partial action by allowing users to correct only the specific keypoints that are erroneous rather than requiring complete manual annotation of all keypoints. The ML model then uses these partial corrections to generate refined predictions for the entire pose, reducing the time investment required while maintaining high accuracy.
Solution Approach 2:
The ML model performs preliminary automated keypoint detection before user correction, providing a head start on the annotation process. This preliminary action gives users a baseline to work from, requiring only corrections rather than complete manual labeling, thereby reducing overall time loss.
3Measurement precision
If iterative ML model processing is used, then measurement precision is improved through correction, but device complexity increases due to multiple processing stages
Solution Approach 1:
The patent employs a universal ML model architecture that can operate in multiple modes: fully automated detection, user-assisted correction, and iterative refinement. The same core model handles all processing stages, reducing the need for separate specialized components and managing complexity through multi-functionality rather than proliferation of separate systems.
Data Source
AI summary
A system includes a hardware processor, a machine learning (ML) model trained to predict two-dimensional (2-D) poses and a graphical user interface (GUI). The hardware processor is configured to receive at least one partial pose input representing a 2-D partial pose of a subject, display, via the GUI, the 2-D partial pose, and receive, via the GUI, at least one user input responsive to the display of the 2-D partial pose. The hardware processor is further configured to predict, using the ML model and in response to receiving the at least one user input, a 2-D full pose of the subject, to provide a predicted 2-D full pose having a plurality of keypoints, and display, via the GUI, the predicted 2-D full pose and the plurality of keypoints.