Scholarlink

Real-time human pose estimation and gesture popularity have been a thrilling topic inside the AI network and features spurred the improvement of progressive methodologies integrating advanced technologies with OpenCV and MediaPipe. This paper aims to develop an AI-powered virtual fitness trainer that tracks different sports and yoga poses, providing real-time feedback to enhance exercise accuracy and safety. This research study uses Google’s Mediapipe library, designed for deep learning and multimodal machine learning pipelines, to demonstrate a virtual fitness trainer. The proposed system leverages posture estimation algorithms to monitor user movements in real time, offering detailed feedback and workout analytics. By utilizing Mediapipe’s advanced algorithms and posture estimation module, this system accurately identifies key body landmarks during workouts, enabling precise movement tracking. Additionally, the system maintains an accurate track of the overall number of reps. By calculating angles and landmarks, the system assesses the user’s performance. This information is then fed into a machine-learning model, which categorizes the exercise’s repetition count and proper posture. By leveraging MediaPipe’s capabilities, this system enables users to improve their form in real-time, making fitness training more accessible, effective, and data-driven.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Fitness with AI: Realtime Pose Estimation Using Deep Learning

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Umesh Gupta,
  • Rajiv Kumar,
  • Kashish Sethi,
  • Jayesh Bayla,
  • Moolchand Sharma

摘要

Real-time human pose estimation and gesture popularity have been a thrilling topic inside the AI network and features spurred the improvement of progressive methodologies integrating advanced technologies with OpenCV and MediaPipe. This paper aims to develop an AI-powered virtual fitness trainer that tracks different sports and yoga poses, providing real-time feedback to enhance exercise accuracy and safety. This research study uses Google’s Mediapipe library, designed for deep learning and multimodal machine learning pipelines, to demonstrate a virtual fitness trainer. The proposed system leverages posture estimation algorithms to monitor user movements in real time, offering detailed feedback and workout analytics. By utilizing Mediapipe’s advanced algorithms and posture estimation module, this system accurately identifies key body landmarks during workouts, enabling precise movement tracking. Additionally, the system maintains an accurate track of the overall number of reps. By calculating angles and landmarks, the system assesses the user’s performance. This information is then fed into a machine-learning model, which categorizes the exercise’s repetition count and proper posture. By leveraging MediaPipe’s capabilities, this system enables users to improve their form in real-time, making fitness training more accessible, effective, and data-driven.