Proper posture during exercise is often the link between achieving you did optimal results and avoiding injuries. This paper introduces a real-time posture correction system for three fundamental exercises: squats, bicep curls, and planks. Mediapipe's pose estimation framework, with the aid of some machine learning techniques, analyzes body movements to detect misalignment. With the flexibility of using either a live webcam feed or pre-recorded video files, users can interact with the system. The ultimate solution identifies key posture deviations such as knee tracking errors in squats, incorrect arm angles in bicep curls, or poor spinal alignment in planks and delivers precise, actionable feedback to help users adjust in real time. The developed tool is intuitive, adaptable, and scalable, making it suitable for individual users and fitness professionals. This work demonstrates the potential of AI-driven systems to enhance workout safety, effectiveness, and accessibility by bridging technology and exercise, by using lightweight AI models and adjustable feedback, the system provides accurate, real-time posture correction.
Paper
Full text
PoseSculpt: Real-Time AI-Driven Posture Aligner and Corrector
Semantic Scholar · 2025
Abstract
Proper posture during exercise is often the link between achieving you did optimal results and avoiding injuries. This paper introduces a real-time posture correction system for three fundamental exercises: squats, bicep curls, and planks. Mediapipe's pose estimation framework, with the aid of some machine learning techniques, analyzes body movements to detect misalignment. With the flexibility of using either a live webcam feed or pre-recorded video files, users can interact with the system. The ultimate solution identifies key posture deviations such as knee tracking errors in squats, incorrect arm angles in bicep curls, or poor spinal alignment in planks and delivers precise, actionable feedback to help users adjust in real time. The developed tool is intuitive, adaptable, and scalable, making it suitable for individual users and fitness professionals. This work demonstrates the potential of AI-driven systems to enhance workout safety, effectiveness, and accessibility by bridging technology and exercise, by using lightweight AI models and adjustable feedback, the system provides accurate, real-time posture correction.