.png){kind=icon}
.png){kind=icon}
Return
Real-time and convenient mirror augmentation framework based on 3D pose estimation application
.png){kind=icon}
PRE
AI
DOI：10.1016/j.dsp.2024.104936.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This article proposes a horizontal pose fusion framework based on mirror images to address the data limitations of single-view three-dimensional (3D) pose estimation and the need for model flexibility. Based on the model's cognitive inertia of dynamic poses, this framework optimizes pose estimation through a mirror method. The method leverages available features from the mirror image of the original image to perform pose fusion estimation using a single-view model. The framework includes a complete theory for motion feature analysis, extraction and utilization outside the model. The problem that needs to be solved belongs to the field of data mining. The flexibility and robustness of the framework can be seen from its design details and experimental performance. Comparative experiments confirm that the method achieves mainstream estimation performance and reaches the level of the latest state-of-the-art (SOTA) method in the field. Finally, we believe it is a real-time, widely applicable, and long-term optimization framework.
Keywords:
Cognitive inertia
Long-term
Mirror pose analysis
Pose fusion framework
Single-view
Journal
IF:
3.6
Papers:
9.9K
Citations:
1.7W
Organization
No organization information available