Web-Based Real-Time Slouch Detection Using PoseNet and TensorFlow.js7
R. Logeshwaran, Gokulakrishnan M, M. Sindhuja · 2025
Prolonged desk work and inadequate ergonomics in remote workspaces have produced substantial health problems linked to posture among students, software developers as well as digital workers and students. Slouching which people perform frequently leads to chronic back pain in addition to spinal misalignment and severe musculoskeletal disorders. Posture monitoring systems based on wearable devices and physical postureadjustment equipment tend to be too expensive while also posing discomfort and being difficult to use outside laboratory settings. The research develops an unobtrusive and instant slouch detection system through browser execution which utilizes PoseNet from TensorFlow.js. The system evaluates slouching posture through webcam body landmark analysis which compares nose position relative to shoulder and ear positions. Users can set how sensitive the system should be for determining slouching through calculation of the normalized vertical distance factor. The system alerts users through desktop notification after recognizing slouching postures. The system differs from conventional AI posture models because it performs its analyses within web browsers on clients' computers with no need for backend operations and no requirements for high-end hardware and app installations.