现代信息科技2024，Vol.8Issue(11)：22-25,4.DOI:10.19850/j.cnki.2096-4706.2024.11.005
基于深度学习的三维人脸重建抗遮挡网络
3D Face Reconstruction Anti-occlusion Network Based on Deep Learning
摘要
Abstract
This paper researches both the face single-occlusion model and the face multiple-occlusion model.It proposes a 3D face reconstruction anti-occlusion network based on Deep Learning,realizing the effective reconstruction of occluded face.The improved single-occlusion model effectively realizes the capture of contextual facial information through pre-training and weight modifications.The improved multi-occlusion model employs a distributed loss function and distinct differentiators to achieve reconstructed facial images through feature distortion and transformation.Experimental results validate that the proposed method can generate more precise 3D facial models across various occlusion scenarios,demonstrating superior robustness and anti-occlusion capabilities.关键词
深度学习/三维人脸重建/单遮挡模块/多遮挡模块Key words
Deep Learning/3D face reconstruction/single-occlusion module/multiple-occlusion module分类
信息技术与安全科学引用本文复制引用
李杏清,王志兵,杨润丰,曾德生,聂影影..基于深度学习的三维人脸重建抗遮挡网络[J].现代信息科技,2024,8(11):22-25,4.基金项目
广东省教育厅2022年度普通高校科研平台特色创新类项目(2022KTSCX385) （2022KTSCX385）
2021年广东省普通高校创新团队项目(2021KCXTD082) （2021KCXTD082）
2022年东莞市社会发展科技面上项目(20221800903482) （20221800903482）
