A 3D Body from Eight Questions – No Photo, No GPU

A questionnaire-based MLP with physics-aware loss predicts 58 body parameters from 8 questions, achieving 0.3cm height accuracy and 3–4cm circumference MAE—comparable to photo-based reconstruction while preserving privacy and running in milliseconds on CPU. The physics-aware loss includes the body model's forward pass to ensure mass and height match user inputs exactly, addressing mass-calculation failures in standard regression. The system serves as a privacy-friendly, cost-effective alternative to traditional photo-based HMR.