Predicting the tactile properties of fabrics from vision and touch

Fabric is one of the biggest markets in the world, as nearly everyone is a consumer of fabrics. Tactile properties of fabrics convey vital information and influence consumer decision and satisfaction. With the rapid development of online shopping, consumers face new challenges, including predicting the tactile properties based on fabric images presented on displays and bridging the gap between visual perception and actual tactile perception. Moreover, the understanding of the tactile properties remains uncomplete due to the multiple tactile properties, the various conditions under which they are assessed, various influencing factors, and the limited efforts made on prediction.
The aim of the present study is to provide a consumer-friendly investigation of the perception of fabric tactile properties. A Leeds Fabric Tactile Database was developed and used in a series of psychophysical experiments to achieve the aim. The database consisted of two parts: Part I included colour-rendered fabric images (flat and draped) along with the corresponding real fabrics, and Part II included real fabric images (flat and draped), fabric rotation videos, and the corresponding real fabrics.
Experiment Phase I was carried out using Part I, evaluating flexible-stiff, smooth-rough, soft-firm, spongy-crisp, and warm-cool under the conditions of flat fabric images, draped fabric images, and touching the fabrics without seeing them (touch-only). The effect of individual fabrics, colour, and experiment conditions was analysed, together with the correlations among experiment conditions and among tactile properties.
Following the experiment Phase I, the experiment Phase II was conducted using database Part II. The experiment conditions were expanded to cover all real-life scenarios of human-fabric interaction, with newly added conditions including fabric rotation videos, viewing the fabrics but not touching them (vision-only using real fabrics), and viewing the fabrics while simultaneously touching them (vision+touch). In addition to the analyses of factors and correlations, predictors were extracted from fabric images and videos to model the perception of fabric tactile properties.
All the psychophysical experiments in the present study applied the method of categorical judgement. The analyses were carried out in alignment with the categorical judgement framework, where each score carried a specific perceptual meaning. Taken together, the present study demonstrated a comprehensive investigation of the perception of fabric tactile properties. By using the fabric images and videos, good prediction can be achieved for the perception of fabric tactile properties.