3D Face Reconstruction Anti-occlusion Network Based on Deep Learning
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.
Deep Learning3D face reconstructionsingle-occlusion modulemultiple-occlusion module
国家科技期刊平台
NSTL
万方数据
