Method And Application Of Sketch-based3D Garment Model Retrieval

With the popularity of the term DIY, personality has become a fashionable pursuit, moreand more people choose to design clothes by themselves. DIY clothing became popularonly in recent years, its technical difficulty is high compared to other hand-made product,including the measurement of body, cutting, sewing and so on. With the development of3D garment CAD technology, the number of3D garment model and garment pattern file isgrowing, providing inspiration for the costume DIY enthusiasts. But a3D clothing modeldoes not correspond to a pattern file on the Internet, and therefore they are not able toproviding tailoring guidance to DIY clothing enthusiasts. Therefore, the development of asimple and effective3D garment model retrieval and tailoring guidance system, has notonly significant meaning to promote the3D garment model reuse, but also has practicaltailoring guidance value for lovers of DIY clothing.Sketch is a natural way of human interaction. So we present a complete sketch-based3Dgarment model retrieval and tailoring guidance system framework. Based on the input ofhand-drawing garment, we provide users with the3D garment models and correspondingpattern files, giving them crop guidance of output model-like garment. The main work ofthis paper is the research and realization of sketch-based3D garment model retrieval andbuilding a dataset that has3D garment corresponding to the garment pattern file, withwhich we can give users tailoring guidance. Firstly, we studied the existing retrievalalgorithm of3D model and implement the extraction of contours and silhouette picturefrom the front view of3D garment model. Secondly, after research and do experiment ofthe main existing image recognition algorithms, we propose a new feature extractalgorithm that based on local Gabor filter bank and improved bag of feature model, ourexperiment shows that our algorithm is effective and robust for sketch recognition.Changes in clothing style are mainly reflected in some of the internal details. Silhouette isnot enough for us to express the variability of clothing styles. So we provide a new partial match retrieval method based on the feature of contours with inner detail information andsilhouette. At last, we build a dataset that has3D garment models and2D garment patternfiles, with them one to one. Based on our experiment and dataset, we develop asketch-based3D garment retrieval and tailoring guidance system. Experiment results showthat our system has a simple input interface and good retrieval result, meet the DIY lovers’need of tailoring guidance greatly.