Development and Applicability Evaluation of an Image-Based BMI Estimation Model for Occupant-Centric Thermal Environment Control

This study develops a face image-based Body Mass Index (BMI) estimation model for occupant-centric thermal environment control and experimentally evaluates its applicability. To overcome the limitations of conventional continuous regression approaches, a categorical classification scheme was introduced and additional Asian facial image data were used for fine-tuning to enhance the model's generalization performance. The proposed model achieved accuracies of 0.9979 and 0.9555 on the training and validation datasets, respectively, and an overall estimation accuracy of 81.8% in experiments conducted under realistic indoor conditions. These results demonstrate that the proposed BMI estimation model can serve as an enabling technology for thermal environment control systems that explicitly account for occupants' BMI.