BROWSE
Detail
Conference Paper
Pose Estimation Score Prediction with Quality Measures of Images
Cited - time in 
Share 
Share
- Authors
- HaeJeong Yu, Dohun Kim, Wonjong Kim
- Issue Date
- 2024-01
- Citation
- International Conference on Green and Human Information Technology (ICGHIT) 2024, pp.1-3
- Language
- English
- Type
- Conference Paper
- Abstract
- Since pose estimation models are primarily sensitive to image quality, preprocessing poor-quality images can enhance their performance. This paper proposes a method to improve the performance of the pose estimation model by utilizing various quality metrics extractable from images (such as brightness, complexity, moments, contrast, etc.) before they are used as inputs for the model. We use regression models to predict pose estimation results using image quality metrics. When applied to the MS COCO validation dataset, the proposed approach achieved a predicted accuracy of 86.73%. These results demonstrate the feasibility of predicting scores for pose estimation models using image quality metrics, suggesting the potential to selectively preprocess images with low scores in the model to enhance overall results.
- KSP Keywords
- Image quality metrics, Pose estimation, Quality measure, Regression Model, estimation model