Posture Correction Using Deep Learning
Fuente:
Zenodo
Guardado en:
| Autores principales: | , |
|---|---|
| Formato: | Recurso digital |
| Publicado: |
Zenodo
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866901751548870656 |
|---|---|
| author | T S, Nandhu Benjamin, Jetty |
| author_facet | T S, Nandhu Benjamin, Jetty |
| contents | <p>Posture is extremely important in keeping oneself physically healthy, enhancing efficiency of exercising, and avoiding injuries. Poor posture at the workouts, yoga, or everyday activities may cause strain in the muscles, joint issues and chronic health problems in the end. In this seminar, a posture correction system derives from deep learning that does real-time analysis of the body posture with the help of a web camera. The system employs the TensorFlow.js to run machine learning models in the web browser, which ensures privacy and low-latency processing.The proposed framework uses the MoveNet pose estimation model that identifies 17 main body joints such as shoulders, elbows, hips, knees, and ankles. Such keypoints create a skeleton of the human body. A geometric pose analogy module estimates trigonometrically the joint angles and compares them with the ideal ranges of postures in a pose library. In case the user has not aligned to the correct range of the posture, the system would provide immediate feedback and posture correction recommendations.The system also has the option to use MediaPipe Pose, which can be used instead as an alternative model that can result in the creation of detailed 3D tracking of body coordinates. The proposed system is effective at monitoring and correcting posture through applying deep learning, computer vision, and geometric analysis. The system can find application in fitness training, physiotherapy, yoga, rehabilitation programs and monitoring of ergonomic posture.</p> |
| format | Recurso digital |
| id | zenodo_https___doi_org_10_5281_zenodo_19559300 |
| institution | Zenodo |
| language | |
| publishDate | 2026 |
| publisher | Zenodo |
| record_format | zenodo |
| spellingShingle | Posture Correction Using Deep Learning T S, Nandhu Benjamin, Jetty Deep Learning, Pose Estimation, Posture Correction, TensorFlow.js, MoveNet, MediaPipe Pose <p>Posture is extremely important in keeping oneself physically healthy, enhancing efficiency of exercising, and avoiding injuries. Poor posture at the workouts, yoga, or everyday activities may cause strain in the muscles, joint issues and chronic health problems in the end. In this seminar, a posture correction system derives from deep learning that does real-time analysis of the body posture with the help of a web camera. The system employs the TensorFlow.js to run machine learning models in the web browser, which ensures privacy and low-latency processing.The proposed framework uses the MoveNet pose estimation model that identifies 17 main body joints such as shoulders, elbows, hips, knees, and ankles. Such keypoints create a skeleton of the human body. A geometric pose analogy module estimates trigonometrically the joint angles and compares them with the ideal ranges of postures in a pose library. In case the user has not aligned to the correct range of the posture, the system would provide immediate feedback and posture correction recommendations.The system also has the option to use MediaPipe Pose, which can be used instead as an alternative model that can result in the creation of detailed 3D tracking of body coordinates. The proposed system is effective at monitoring and correcting posture through applying deep learning, computer vision, and geometric analysis. The system can find application in fitness training, physiotherapy, yoga, rehabilitation programs and monitoring of ergonomic posture.</p> |
| title | Posture Correction Using Deep Learning |
| topic | Deep Learning, Pose Estimation, Posture Correction, TensorFlow.js, MoveNet, MediaPipe Pose |
| url | https://doi.org/10.5281/zenodo.19559300 |