Facial expression is an important information carrier for communication and display,and 3D face reconstruction plays a vital role in image recognition,animation,film and other fields.Aiming at the problems of losing details and high computational complexity in 3D face reconstruction,a 3D face reconstruction design based on 2D view is proposed.Based on two-dimensional view,the method proposed in this paper uses Fourier neural operator to solve Poisson equation,and reconstructs the mesh from directional point cloud measurement,which improves the limitation of classical depth neural network in shape reconstruction and obtains good results in reconstruction quality and running time.In addition,in order to quantify facial expression recognition,a measurement method based on vertex distance is proposed,and the expression state is classified by support vector machine.
3D face reconstructiontwo-dimensional viewfourier neural operatorvertex distance
NETL
NSTL
万方数据
