Enhancing Keypoint Detection Accuracy Using Pose Estimation in Pressure Images
We extracted keypoints from pressure images with less information compared to typical RGB images. Therefore, the initial keypoint extraction results are not accurate. To improve accuracy, we estimated the pose and then applied post-processing based on the pose estimation results. The post-processing involved calculating the average value of keypoints for each pose and correcting the keypoints with low extraction scores by replacing them with these average values, thereby enhancing the overall accuracy of the keypoints extraction.
Authors
- Seihyoung Lee
- Yun ji Ban
- Hongyeon Yu
- Jiwon Park (ORCID: https://orcid.org/0009-0000-4323-5684)
Institutions
- Electronics and Telecommunications Research Institute (KR)
Publication Details
- Published
- 2024-10-16
- DOI
- https://doi.org/10.1109/ictc62082.2024.10827473
- Primary Topic
- Image and Object Detection Techniques
- Type
- article
- Field-Weighted Citation Impact
- 0.00