Abstract
Due to the global focus on sustainability in recent years, reducing the mass waste of clothing has become one of the important issues. A wearing simulator is thought one solution to such an issue because there is no need to stock large quantities of clothing in stores for trying on. However, conventional wearing simulators cannot accurately reproduce knitted products because prediction of their behavior at the stitch level is very time-consuming. Therefore, in this study, a method is proposed to predict the knitted fabric shape at the stitch level based on deep learning. To use deep learning, it is necessary to prepare large and diverse datasets for training. So, a procedure to efficiently generate such datasets is proposed considering the characteristics of stitches in deformed fabrics at first. Next, a shape predictor of each stitch is constructed with neural networks and learned. Then, the predictor can derive the stitch shape immediately when the boundary conditions of the stitch are given. Finally, by optimizing the boundary conditions of all stitches in a knitted fabric, the deformed shape of the fabric at the stitch level can be predicted.
