Out of Home (OOH) advertising models through deep learning method with demographical information such as gender, age, etc. While a more comprehensive model would involve fine-tuned information from audience. This paper proposed a subdivided apparel recognition model to enhance the existing audience measurement for OOH. SVM accompanied by Libra RCNN and histogram intersection kernels is adopted alongside advertising board-mounted cameras, which obtain unprocessed data from which gender, age and other demographic features are discerned to determine viewers of particularly clothing advertising. Pervasive adoption for contactless consumer engagement, customised content display and consumer analysis is possible through the amalgamation of results, while audience measurement via digital advertising panels can be more effectively understood by OOH companies and businesses.