BROWSE
Detail
Conference Paper
연관관절의 혼합 히트맵을 이용한 Pose Estimation 개선방안 연구
Cited - time in 
Share 
Share
- Issue Date
- 2024-06
- Citation
- 대한전자공학회 학술 대회 (하계) 2024, pp.2557-2560
- Publisher
- 대한전자공학회
- Language
- Korean
- Type
- Conference Paper
- Abstract
- In this study, we propose a novel heatmap fusion technique to enhance pose estimation using the information of relation between joints. Unlike conventional approaches, which rely on independent Gaussian heatmaps, our method incorporates the distribution of neighboring joints to improve the performance of pose estimation. Experimental results demonstrate that it can find better results for challenging images where the model have low confidence scores for some joints. By leveraging the joint distribution information, our method offers a more flexible heatmap representation, leading to better pose estimations results in difficult cases. These findings underscore the importance of considering joint interactions in pose estimation models and highlight the potential of our approach to improve performance in specific scenarios.
- KSP Keywords
- Estimation model, Pose estimation, distribution information, fusion technique, joint distribution