Real-time three-dimensional human body posture estimation method and device for spatio-temporal topology modeling
Through the method of space-time topology modeling, using contrast learning and lightweight spatio-time human body topology extraction network, the problem of difficulty in deploying human body pose estimation algorithm on edge devices is solved, and efficient and accurate three-dimensional human body pose estimation is achieved.
Patent Information
- Application Number
- CN202411591535.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-08
- Publication Date
- 2025-03-28
- Estimated Expiration
- 2044-11-08
AI Technical Summary
On edge devices with limited computing resources, it is difficult to deploy human posture estimation calculation methods and insufficient extraction of human body topology affects the accuracy and efficiency of three-dimensional human posture estimation.
The lightweight three-dimensional human body pose estimation method is adopted for space-time topology modeling. Label-free pre-training is performed through comparative learning paradigm, and a priori information of the human body topology structure is extracted, and a lightweight space-time human body topology extraction network is used to extract the human body topology relationship between the time domain and the space domain in parallel, and the three-dimensional skeleton coordinates are output through linear regression.
It realizes efficient and accurate three-dimensional human pose estimation on resource-constrained edge devices, improves computing efficiency and pose estimation accuracy, and has good generalization ability and real-time reasoning performance.