using 2D pic to learn the facial features then map them to 3D mesh model, extracting the feature from vertex nor. Also trying curvature filter & cluster and spheric harmonics way
This project is built by vs2015.
Libs:
use opengl, boost, eigen3. You must download the glut , the boost and the eigen3 for drawing, math and matrix calculation.
Files:
in 2d_feature dir, there are 11 3D face.
mesh.bin ，stores the triangle mesh model with structure :
nv(an integer ，describes the number of vertexs)
nf(an integer ，describes the number of meshs)
vb(nv * 3 floats ，describe the coordinate of the nv vertex)
ib(nf * 3 integers ，describe the vertex index of each triangle mesh)
marks.txt :
the feature points learned from 2D pic(no 2D pic upload cause the privacy problem)
从二维平面图中提取的特征点在三维空间上对应的坐标
sharp_point.txt :
the coordinate of the vertexs which pass curvature check
满足曲率一定大小的点的坐标
DBScan_result.txt :
clustering the sharp_points by DBScan, then find the center of each classes. This file stores the coordinate of the centers and something for drawing the model.
对过滤后的点用DBscan方法进行聚类，这个文件存储了这些中心的坐标以及一些用于绘图的数据
DBScan_feature.txt :
mark the center as eyes, ears, nose tip artificially. the variance of the vertex normals of the vertexs around the facial features are the final feature of a 3D face.
人为地对聚类后的点标注为眼睛，耳朵，鼻头。以这些特征点的邻域的点的法向量方差作为最后的特征点。
spheric_harmonic_feature.txt ：
the method refers to the paper "Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors" (not sure about the corectness)
采用文章《Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors》的方法（不能保证其正确性）
feature_from_2d.txt :
the variance of the vertex normals of the vertexs around the 2D facial features are the final feature of a 3D face.
利用2D特征点周围的点的法向量方差作为特征点
in dir test, there are the code and the project files(Mesh.h cpp, drawface.cpp, Vertex.h cpp , the 5 files are useless)