Toward Smart Internet of Things (IoT) for Apparel Retail Industry: Automatic Customer's Body Measurements and Size Recommendation System using Computer Vision Techniques

Internet of Things (IoT) is not a buzzword in the media outlets anymore. Over the last five years, many products, that are defined as IoT devices are widely offered for consumers starting from embedded and portable sensors with Internet connectivity to remotely distributed systems with machine learning (ML)-based-intelligence and cloud connectivity. We believe that IoT devices will widely spread as reliable products if and only if they are equipped with efficient human computer interaction (HCI) features, specifically visual interaction with users based on computer vision and machine learning. An interesting applied field for such devices is the apparel retail industry where outlets and consumers still depend on legacy methods for fitting, which cost the industry hundreds of millions due to damage or return goods. Therefore, we propose a ML-based human body measuring and size recommendation system that can provide customers of online and retail stores with accurate estimates of their body measurements with respect to the retailer size charts. To the best of our knowledge, this is the first system to automatically measure body parts and predict subject's size using two RGB images only. In addition, we developed a new body measurements dataset from scratch due to the unavailable public access to such dataset. With an average error range from 0.5 to 2.5 cm, our system shows promising results for seven standard body measurements used for cloths size suggestion. In this paper, we focus on providing an innovative and new experience to revolutionize apparel industry and promote well being.