生成对抗网络在服装图片遮挡修复上的应用

The Application of Generative Adversarial Networks in the Restoration of Occluded Areas in Clothing Images

纺织品属于有机物质,其化学稳定性比较差。纺织品文物出土后经常会出现褪色、残缺、破裂、污损等问题,这严重影响了古代服饰的考究。传统的纺织品修复大都采用人工修复的方法,但人工修复费时费力,且修复效果因人而异。基于此,文章用遮挡的服装图片模拟残缺的出土服装,应用生成对抗网络对其进行修复,分析生成对抗网络的修复效果,并建立古代服装数据集,为今后的服装修复研究提供新的思路和途径。

Textiles belong to organic substances,with relatively poor chemical stability.Textile relics often suffer from issues such as fading,incompleteness,breakage,and contamination after excavation,greatly affecting the meticulous study of ancient costumes.Traditional textile restoration methods mostly rely on manual intervention,which is time-consuming and laborious,with restoration results varying from person to person.Therefore,the article simulates damaged excavated garments using obscured clothing images and employs Generative Adversarial Networks for restoration.It analyzes the restoration effects of Generative Adversarial Networks,establishes a dataset of ancient clothing,and provides new perspectives and avenues for future research in garment restoration.