由LeNet-5从单张着装图像重建三维人体

提出基于LeNet-5的从单张着装图像恢复人体三维形状的方法,建立着装人体正面轮廓和人体形状空间之间的映射模型,实现了高效、精确的三维人体建模,可以应用于对人体表面形状精度要求较高的场合,如虚拟试衣.基于PGA在流型空间上对公开的三维人体数据集进行数据扩增,给虚拟人体进行着装,构建着装人体数据库.从着装人体正面投影图像中提取信息,以人体形状参数及正、侧面轮廓信息为约束,基于LeNet-5完成三维人体重建.实验证明,对于身穿不同款式服装的人,采用的模型通常都能从单张着装图像中重建得到较高精度的三维人体模型.

复杂场景下的人体轮廓提取及尺寸测量

自然拍摄的人体照片由于背景图案较为复杂,采用传统基于图片色彩空间或能量梯度的图像处理方法难以准确地识别人体的轮廓。采用神经网络的方法,可以提高识别的精度。但是,一般的神经网络方法由于计算量与参数规模较大,难以在移动终端部署。因此,提出了一种轻量级的神经网络策略以提取人体轮廓。该网络采用MobileNet V2与U-Net框架,通过构建特定姿态的人体数据集进行训练,识别相应的人体轮廓形状。人体轮廓经过提取关键点、拟合回归分析等后续处理,可估算人体的尺寸。该方法可应用在移动终端上,通过拍摄的人体照片的方法测量人体的尺寸。实验表明,该方法能准确地提取复杂背景照片中的人体轮廓并测量尺寸,在速度与存储占用方面较一般神经网络有一定优势。

Pattern design on 3D triangular garment surfaces

This paper focuses on a pattern design method for a 3D triangular garment surface. Firstly, some definitions of 3D style lines are proposed for designing the boundaries of patterns as drawing straight lines or splines on the triangular surface. Additionally some commonly used style lines are automatically generated to enhance design efficiency. Secondly, after style lines are preprocessed, a searching method is presented for quickly obtaining the boundaries and patches of a pattern on the 3D trian- gular surface. Finally a new pattern design reuse method is introduced by encoding/decoding the style line information. After style lines are encoded, the pattern design information can be saved in a pattern template and when decoding this template on a new garment surface, it automates the pattern generation for made-to-measure apparel products.