The inherent brittleness of dried bacterial cellulose (BC) films significantly restricts their use in the textile industry. This study addresses this limitation by developing flexible, sewable BC sheets via an environmentally and human-friendly modification. The material was prepared from Acetobacter xylinum NQ-5–derived BC modified with polyethylene glycol (PEG) and plant-based edible oils (a food-grade strategy). Structural analyses (XRD and FTIR) confirmed that this simple modification effectively reduced crystallinity and attenuated the rigidifying hydrogen-bonding network within the BC matrix. Mechanically, the synergistic addition of PEG and coconut oil yielded the lowest bending rigidity (84.2 mg·cm), the highest crease recovery (66.7%), and a drastic ∼91% reduction in Young’s modulus (51.2 MPa) compared to the untreated BC Control. Notably, the modified material achieved a seam stress of 4.67 MPa, reaching a level comparable to standard cotton fabric, whereas the brittle untreated BC Control failed at 1.93 MPa. These structural and mechanical improvements enabled the BC sheets to successfully withstand continuous stitching using a household sewing machine without rupture. Demonstrating practical utility, a bustier prototype was constructed using the modified BC sheets, entirely free of tearing or structural failure during the sewing process. In summary, this study presents a safe, sustainable, and scalable approach that transforms rigid BC into a flexible, truly sewable textile substrate, marking a novel and practical advance toward its application in sustainable apparel and other lifestyle materials.

In the manufacturing of synthetic fibers, uniformity of the fiber properties is a critical quality control parameter. Currently, visual assessment by expert inspectors is the predominant method for evaluating the uniformity of synthetic fiber. A quantitative evaluation method using image analysis was investigated to establish a more accurate and efficient approach. As a fundamental step, this study focused on quantifying the evaluation scale of limit samples. The limit sample was a dye-circular plain-knitted fabric comprising five units, each of which was assigned a grade of fiber or yarn quality. The uniformity of synthetic fiber properties, especially the degree of streaky dyeing unevenness observed in dye-circular plain-knitted fabrics, was demonstrated to be expressed using image features. The image features were statistics about the gray-level gradients in the wale direction, obtained by horizontal edge detection on an image with loops removed by low-pass filtering. Furthermore, these image features and grade values representing synthetic fiber uniformity exhibited a linear relationship with a remarkably high coefficient of determination.

This paper introduces FabricDefectNet, an AI-driven visual inspection system for automating weaving inspection in the textile industry. Conventional AI inspection systems require many defective sample images for training, and the cost of data collection is a significant barrier to practical application, especially in today’s high-mix, low-volume production environment. To solve this problem, our system adopts a novel approach that uses fabric simulation images (generated from design data) not as direct training data, but as reference templates for comparison with the actual fabric. The core of the proposed system consists of three stages: (1) robust feature matching between the simulation image and the real fabric image using the state-of-the-art LightGlue matcher; (2) precise correction of non-rigid deformations (stretching, distortion, skew) in the fabric image using Thin Plate Spline (TPS) warping; and (3) pixel-wise anomaly detection by measuring cosine distances between the aligned patches. Evaluation experiments with a prototype system demonstrate that FabricDefectNet can accurately detect major defects such as color mismatches, pattern errors, and pitch differences without using any actual defect samples. This work demonstrates a practical AI tool that can detect unknown defects using only normal (design) data, which is expected to improve the efficiency and consistency of quality control in textile manufacturing.

Quality control of synthetic fibers currently depends primarily on the visual evaluation of streaky uneven dyeing in knitted fabrics by expert inspectors, a method limited by its lack of quantifiability and objectivity. As a quantitative alternative, a previously proposed method utilizing the mean gray-level gradient from image analysis proved effective for a limited set of samples; however, it failed to generalize to routine samples exhibiting diverse defects. To address this limitation, this study identified image features capable of objectively detecting uneven streaky dyeing in routine samples. An analysis of inspector markings revealed significant variability in human judgment while also indicating the importance of gray-level gradient patterns with relatively high contrast. By incorporating the latter with geometric characteristics, a new feature was defined: a region displaying a gray-level gradient value exceeding 25% of the upper limit in conjunction with an elongated shape characterized by an aspect ratio of at least 7.0. Validation experiments demonstrated that this feature achieved a high detection rate (recall) of 92.5% for inspector-identified instances. These findings establish a foundation for comprehensive evaluation based on the intensity, length, and number of individual streaks, representing a significant step toward fully automated quality control.