3D Face Reconstruction Based on Channel Multi-scale Feature Fusion
WANG Chaozhi
ZHU Lei
LIU Qingshan
Abstract：In order to solve the problem of lack of shape details of 3D face reconstruction,an unsupervised 3D face reconstruc-tion method based on channel multi-scale feature fusion is proposed in this paper.The channel multi-scale feature fusion method fuses the feature information of different channels on the multi-channel,and learns the facial features of different scales,which makes the network learning richer global features.At the same time,the latest large kernel attention method is introduced to sup-press the background noise on a single channel and extract more human face feature information.Qualitative and quantitative experi-mental results on three face datasets BFM,3DFAW and Photoface demonstrate the effectiveness and superiority of the proposed method.
Keywords：3D face reconstructionchannel multi-scalelarge kernel attentionself-supervision
Publication Date：2025-04-20
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：7( 1002-1007,1014 )