Control and optimization of human perception on virtual garment products by learning from experimental data

This paper proposes an original study for controlling, in an Internet and virtual reality-based collaborative design platform, human perception on 3D virtual garments by adjusting fabric parameters, constituting the inputs to the garment CAD software. This study will permit the designer to determine the most relevant product through a number of interactions with the consumer. For this purpose, two sensory experiments have been realized on a small number of fabric samples. In the first experiment, we propose an active learning-based experimental design in order to find the most appropriate values of the fabric technical parameters permitting to minimize the overall perceptual difference between real and virtual fabrics in static and dynamic scenarios. The second sensory experiment aims to extract normalized tactile and visual sensory descriptors characterizing human perception on the concerned fabric samples. In the collaborative design process, these normalized descriptors will be used for communications between the designer and the consumer on perceptual quality of virtual products. By learning from the experimental data on identified inputs (fabric parameters of the CAD software) and outputs (sensory descriptors), we model the relationship between fabric technical parameters and human perception on finished virtual products. The method of fuzzy ID3 decision tree has successfully been applied in this modeling procedure. This model, combined with the corresponding garment CAD software and the learning data acquired from the two sensory experiments, constitutes the main components of the learning data-driven collaborative design platform. Using this platform, we have realized a number of garments meeting consumer’s personalized perceptual requirements.