Efficient 3D Face Reconstruction Model Based on Dense Mesh Solution Using Rendering and Partial Search
Tran Duc Long, Ngo Hai Linh · 2021
One of the most difficult aspects of digitizing the human body is reconstructing a human face model from images. Many independent studies on face modeling methods have been published; however, the majority of the studies focus on improving the accuracy of the generated face model rather than the ability to integrate facial models with body models to create a complete human model. As a necessary consequence, the authors implement a method capable of effectively and accurately integrating with body models based on Position Map Regression Network (PRN) - a point cloud generation from an image based on a convolutional neural network (CNN).