Assessment and Semantic Categorization of Fabric Visual Texture Preferences

1.1. Toward E-commerce: Visual Perception of Fabric Structure

The online shopping experience encompasses several factors, such as shortening shopping time, facilitating price comparison, access to product and technology news, and others. However, some aspects of e-commerce regarding the experience of contact with the product are a hindrance [23]. Lester at al. [24] argue that one of the major shortcomings of online shopping is that shoppers cannot touch the products. Citrin at al. [25] explain that the need to touch products is negatively related to online purchasing, especially for clothing. The shopping experience (and buying decision) can be impaired because of the desire to try on, touch, and feel the clothes [26]. Since apparel may pose certain risks when purchased online, it is important that e-companies develop strategies and solutions to reduce this risk. Therefore, understanding the differences between consumers and their knowledge and perceptions regarding the appealing product and fabrics can consequently maximize purchase intention.

The e-commerce fashion industry is on the rise and everything is sold based on visual perception. However, while the absolute numbers of clothing items are steadily increasing, one of the biggest problems of the e-commerce industry is the lack of opportunity to physically try on and touch a product. To compensate for this disadvantage, online retailers need to sell an experience, not just a product. We reviewed the literature and found that there is a lack of in-depth studies on the visual perception of fabric textures. There are studies on predicting the subjective feel of fabrics and the perception of fabric softness [27, 28], a study on the subjective feel of leather in automobiles [29], and a study on the expression and verbalization of tactile feelings [30].

Zhao et al. [31] set up an online questionnaire in which they asked participants to view a dynamic apparel picture and select their preferred words to describe the picture. Moody, Kinderman, and Sinha [32] conducted an experiment to investigate clothing preference by asking participants to view different clothing styles and then select their preferred ones. In another study, Zhao et al. [33] investigated summer shirt fabric preferences and their characteristics—the influence of gender, visual, and tactile influence on fabric preference. They found that visual details were essential factors influencing customer preferences.

When we see an object, we generally have a good idea of how it would feel [34, 35]. This visual prediction of haptic material properties has practical implications for retail businesses that rely on visual communication [36]. It is well known that an appealing visual display of products increases consumer purchase intention [37, 38]. Napompech [39] argues that online apparel sellers should post realistic images of garments from different angles on their website to help buyers imagine how the garment would fit on their own bodies. Wijntjes et al. [36] found that dynamic visual information (a movie) reveals more about how fabrics feel than static visual information (a picture). Information such as color, size, and fit were found to be particularly important factors in reducing perceived risk in clothing purchases, especially when they provided information about sensory and experiential aspects [38]. We all have a variety of memories about textiles that come from experiences of looking at, touching, and wearing them [40].

In 2015, Shanthi and Kannaiah [41] noted that online shopping is a growing reality that helps consumers reduce search time for products and services, and today online shopping is booming—a pandemic year drove many more shoppers online. According to a Statista report [42], fashion is the largest market segment in B2C eCommerce, and its global size was estimated to be $525.1 billion in 2019. The market is expected to continue growing at 11.4% per year [43]. In previous studies, researchers have used various techniques to measure visual texture preferences, and Kansei engineering has been widely used in this context. Kansei engineering is a consumer-oriented technology for new product development that aims to develop or improve products and services by transferring customers’ psychological feelings and needs into the product design domain [44, 45]. Often users use Kansei words or adjectives to describe their feelings about products. Therefore, it is important how products are explained semantically. Küller [47] has developed a structured method called semantic environment description (SMB). The SMB method is easy to administer, has high reliability and validity, can be easily adapted for cross-cultural comparisons, and provides useful knowledge for design, engineering, and marketing.

The semantic environment description (SMB method) is a method that can be used to evaluate and better understand the overall impression of an environment, and according to Küller [46] it can help to reveal how people assess their impression based on pleasantness and the information rate. Küller's adjectives were originally grouped into eight factors: Pleasantness, Complexity, Unity, Enclosedness, Social Status, Potency, Affection, and Originality [47]. These factors and KWs have been used in studies related to computer-based environment [48], indoor environment with emphasis on office interior [49, 50], virtual environment [51], outdoor environment at nursing homes [52], healthcare environment [53], the automotive industry [54, 55], e-commerce [56], and many other fields. Lokman and Kamaruddin [57] proposed affinity clusters for affective product design where words are collected for each topic.

However, we believe that the visual texture of fabrics should be explained semantically, and the aim of this study is to define the consistency of kansei words or adjectives that best describe the affective characteristics of clothing products. Our interest was to find out how the texture of fabrics and their material composition is perceived and semantically described by users. This will help both industry and academia to capture users’ subjective emotional experiences of fabric texture preferences.

2.1. Materials—Collection of Adjectives

The first step of this study was to extract adjectives that describe the impression one has of a texture, such as attractive, luxurious, dynamic, etc. (see all adjectives in Table 1). In this study, 34 adjectives (17 pairs) were collected from WGSN, experts and pertinent literature. Based on these adjectives, a checklist was developed to assess the samples in the evaluation session. The checklist includes all adjectives arranged on a 7-point Semantic Differential scale from a negative to a positive connotation.

The search procedure for the adjectives was based on a review of the literature—scientific papers, but also fashion articles. We looked for adjectives suitable for the description of the visual texture of fabric. We paired the appropriate adjectives with their positive/negative poles. After extracting the adjectives, we checked their distribution in research from different areas (see Table 1).

2.2. Materials—Collection of Fabric Visual Texture Database of Stimuli (Samples)

The visual texture database of stimuli used in our experiment consists of 12 types of commercially available fabrics with different patterns, as shown in Figure 1, and they were used as samples. All samples were mostly beige or white, but they were converted into binary (black and white) images.

Figure 1.

Fabric texture samples.

Selected samples were produced by various manufacturing technologies: weaving, knitting, knitting finished with fulling to achieve a particularly soft look and handle, and leather processing. The raw materials used to produce the selected textures were either natural, such as cotton, wool, silk, and leather/suede, or man-made such as polyamide, polyester, polyacrylonitrile and polypropylene. The influential factors that define the textures are summarized in Table 2.

The selected textures are described below.

Jersey is a soft, stretchy, knitted fabric originally made of wool, but now also made of cotton, blends, and synthetic fibers. Single jersey is a lightweight fabric made on a single knitting machine needlebed [58, 59]. Its technical face is smooth, with the side limbs of the needle loops having the appearance of columns of Vs in the wales. On the technical back, the loops form columns of interlocking semicircles [60]. Jersey stretches slightly more in the transverse direction than in the longitudinal direction. It also tends to curl at the edges and is less stable than some other types of knits [61].

Satin weave has a visible sheen and feels smooth, which is due to a tightly woven structure that allows the yarn to lay across the surface of the fabric. Fabrics in satin weave are often used as linings because they glide easily over other garments. They typically have a glossy surface and a dull back and drape well. The compactness gives the satin fabric more body and less porosity, making the fabric warmer. Long yarn floats make it susceptible to abrasion and snagging [61, 62].

Faux (fake) fur is a type of textile fabric made to simulate genuine animal fur. It is a pile fabric and is typically made from polymer fibers that are processed, dyed, and cut to match a specific fur texture and color. The primary polymer used for faux fur is acrylic and modacrylic. Today's faux furs are almost indistinguishable from the natural furs they imitate. They are more durable and resistant [63, 64].

Leather is a material made by tanning animal hides or skins. Leather stretches but does not return to its original shape. It is one of the oldest fabrics used by humans for clothing. It is not constructed but cut from the animal as a skin [58].

Canvas is an extremely durable, heavy, tightly woven, and very stiff fabric made from a uniform yarn. It was formerly used for sails and was made in a tight plain weave [58, 61, 65]. Today it is used for sports equipment and footwear, sportswear, workwear, bags, upholstery, art supplies, etc. [66].

Jacquard is a woven structure that creates patterns and textures through a complicated weave system in which warp and weft threads are raised and left. An intricate pattern is woven on a special mechanical loom rather than printed on the surface. Compared with printed designs, a jacquard pattern is much more wearable and retains its luster over a longer period of time. Cloqué is a lightweight fabric with a puckered, sculpted, or bubbled effect on the surface. It has a high/low relief pattern with shiny and matte areas. The ruffle effect is created by weaving yarns under varying tension. Cloqué jacquard is a fabric with a jacquard weave that has a ruffle or bubble effect. It is often made from silk, wool, and cotton blends [58, 65, 67, 68].

Nylon (polyamide) is a strong, lightweight synthetic fiber, but it melts easily at high temperatures. It is also a smooth fiber, which means that dirt does not easily adhere to its surface. It has very low absorbency, so it dries quickly and does not require ironing. Nylon has long been of major importance in the manufacture of women's hosiery as it can be heat set and is strong, elastic and abrasion resistant. In the filamentous form, it is characterized by moderate sheen. The main advantage of polyamide fabric is its elasticity and softness. It does not retain heat effectively or transport moisture to the same extent as other fabrics, but it can be formed into thin fabrics such as stockings [61, 67, 69].

Suede is a type of leather made from the underside of animal skin, giving it a soft surface. It has a fuzzy surface produced by brushing the flesh side of the hide with abrasive-coated rollers. It is usually made from lambskin, but can also be made from other types of animals, including goats, pigs, calves and deer. Suede is softer, thinner, and not as strong as full-grain traditional leather [58, 61].

Terry is a fabric with loops that can absorb large amounts of water. It is mostly used for towels and bathrobes. It can be made by weaving or knitting. It is usually made of 100% cotton, but can sometimes contain polyester. In the fabric, the ground is visible in the spaces between the uncut loops of the pile. Terry has a moderately soft drape. The loop pile tends to snag and pull. The fabric gives off lint easily [61, 62].

Boiled wool is made by boiling woolen fabrics to shrink and compress the fibers. It is a thick and dense knitted fabric with the stability and shape retention of a woven fabric. It is very durable, comfortable to wear and warms in inclement weather conditions. It looks the same on both sides [70].

Ribbing is the general term for any knitting pattern that results in vertical columns of knit and purl loops. Knitted rib is a knitted fabric with alternating raised and lowered rows. It is more elastic and durable than plain knitted fabric [58].

Crochet is a process of making textiles by intertwining loops of yarn, thread, or strands of other materials with a crochet hook. The word crochet comes from the French word croc, meaning “hook.” Stitches are made with a single crochet hook to pull one or more loops through the previous loops of a chain. This construction can be built up to create a patterned fabric [58].

2.3. Procedure and Image Acquisition

Subjects were asked to complete an online questionnaire. Then, 12 samples (fabric texture stimuli) were shown one by one with 34 adjectives (17 pairs). The arrangement of the samples (see Figure 2 and Table 1) is numbered as presented in the study. The online survey system/tool did not allow us to randomly arrange the fabric texture samples. Subjects were asked to rate their feelings in the checklist according to the scale provided. The questionnaire included two sections: (1) demographic data of the respondents, including age, gender, education, and occupation; and (2) sensory experience, consisting of 204 questions. Images were presented as rectangles (2,400 × 700 pixels) on a white background. All respondents were required to adjust their screen settings to maximize the brightness of the screen, and this was a prerequisite to begin completing the questionnaire. Images were captured using a Canon IXUS 50 camera (lens model: 5.8–17.4 mm; focal length: 5.8 mm). The specimens were illuminated by a natural (sun) rather than an artificial light source. The camera was placed perpendicularly above the sample surface, the original image size was 2,592 × 1,944 pixels, and the image resolution was 180 dpi. Finally, the region of the image was set to 2,400 × 700 pixels and the image resolution to 180 dpi.

Figure 2.

Dendrogram using Ward's method.

Figure 3.

Selected textures: satin and leather.

Figure 4.

Selected textures: suede and boiled wool (top row); faux fur and terry (middle row); cloqué jacquard and crochet (bottom row).

Figure 5.

Selected textures: jersey and nylon (top row); knitted rib and canvas (bottom row).

2.4. Data Analysis

HCA is a cluster analysis method that aims to form a hierarchy of clusters. Using this method, we were able to identify an average linkage between the groups consistent with the textile samples. For the clustering method, we used Ward's method to measure the squared Euclidean distance. HCA was performed with the aim of determining the number of clusters and their consistency. We then performed k-means clustering with the selection of three main semantic categories. The aim was to find out which pairs of adjectives are grouped together and how they are related to the textile samples.

As an alternative approach to the same goal—grouping adjective pairs—we adopted another multivariate method, PCA. PCA creates new variables, known as principal components (PCs), which are orthogonal linear combinations of the original variables. The number of extracted PCs is equal to the number of variables, but often the first few PCs already account for most of the variability in the data, so the higher ones mainly represent data noise and can be neglected. A thorough mathematical description of the procedure can be found in Jolliffe [71].