Development an Image Recognition-based Clothing Estimation Model for Comfortable Building Thermal Controls

The purpose of this study is to develop an intelligent model that can estimate the clothing insulation (CLO) of occupants using real-time images. Also, performance and applicability of the model to the actual environment were analyzed through the experiment. A total of 16 individual garments and 9 clothing ensembles were set for the model development. The model was developed using the YOLOv5 network and trained on the collected clothing data. The classification performance of the developed model was denoted as 86.7% on average. The applicability of the model was evaluated using the real-time images of the subjects in the test-bed. As a result, the insulation value of the clothing ensembles can be accurately estimated with the MAE of 0.01 clo. This study confirmed the outstanding performance of the CLO estimation model and its high applicability to the actual indoor environment. Therefore, employing the CLO estimation model can contribute to improvement of occupant’s thermal comfort, and it is expected to be applied to various systems capable of PMV-based control.