摘要
针对三维重建方法计算量大,耗费内存高的问题,提出了一种基于知识蒸馏的人体三维模型重建方法应用于电子商务服装定制,利用Resnet-18构建人体正面、侧面掩膜数据集,分别针对男性、女性、中性三种人体特征提取信息,然后通过利用知识蒸馏的方法进行模型压缩,使其能够部署于移动设备。实验表明,论文方法能够在移动设备上以较快的速度、消耗消耗较少的储存空间,实现三维人体模型的重建。
Aiming at the problems of large amount of calculation and high memory consumption of 3D reconstruction method,a 3D human model reconstruction method based on knowledge distillation is proposed,which is applied to e-commerce clothing cus⁃tomization.Resnet-18 is used to construct human front and side mask data sets,and extract information for male,female and neu⁃tral human features respectively,then the model is compressed by using the method of knowledge distillation,so that it can be de⁃ployed in mobile devices.Experiments show that this method can realize the reconstruction of three-dimensional human model on mobile devices with faster speed and less storage space.
作者
万乐玲
行鸿彦
王水璋
张晋宝
WAN Leling;XING Hongyan;WANG Suizhang;ZHANG Jinbao(Collaborative Innovation Center for Meteorological Disaster Prediction and Evaluation,Nanjing University of Information Science and Technology,Nanjing 210044;Jiangsu Key Laboratory of Meteorological Detection and Information Processing,Nanjing University of Information Science and Technology,Nanjing 210044;Nanjing Institute of Multi-platform Observation Technology,Nanjing 214044;Shanxi Yirenhui Technology Co.,Ltd.,Taiyuan 030000)
出处
《计算机与数字工程》
2023年第11期2650-2653,2723,共5页
Computer & Digital Engineering
基金
国家重点研发计划项目(编号:2021YFE0105500)
国家自然科学基金项目(编号:62171228)资助。
关键词
三维人体重建
知识蒸馏
移动设备
人体掩膜
3D human body reconstruction
knowledge distillation
mobile equipment
human mask
