问题描述
我一直在试图推行变形目标动画在OpenGL与面部Blendshapes但以下教程。动画的顶点着色器看起来是这样的：
I've been trying to implement Morph Target animation in OpenGL with Facial Blendshapes but following this tutorial. The vertex shader for the animation looks something like this:
#version 400 core
in vec3 vNeutral;
in vec3 vSmile_L;
in vec3 nNeutral;
in vec3 nSmile_L;
in vec3 vSmile_R;
in vec3 nSmile_R;
uniform float left;
uniform float right;
uniform float top;
uniform float bottom;
uniform float near;
uniform float far;
uniform vec3 cameraPosition;
uniform vec3 lookAtPosition;
uniform vec3 upVector;
uniform vec4 lightPosition;
out vec3 lPos;
out vec3 vPos;
out vec3 vNorm;
uniform vec3 pos;
uniform vec3 size;
uniform mat4 quaternion;
uniform float smile_w;
void main(){
//float smile_l_w = 0.9;
float neutral_w = 1 - 2 * smile_w;
clamp(neutral_w, 0.0, 1.0);
vec3 vPosition = neutral_w * vNeutral + smile_w * vSmile_L + smile_w * vSmile_R;
vec3 vNormal = neutral_w * nNeutral + smile_w * nSmile_L + smile_w * nSmile_R;
//vec3 vPosition = vNeutral + (vSmile_L - vNeutral) * smile_w;
//vec3 vNormal = nNeutral + (nSmile_L - nNeutral) * smile_w;
normalize(vPosition);
normalize(vNormal);
mat4 translate = mat4(1.0, 0.0, 0.0, 0.0,
0.0, 1.0, 0.0, 0.0,
0.0, 0.0, 1.0, 0.0,
pos.x, pos.y, pos.z, 1.0);
mat4 scale = mat4(size.x, 0.0, 0.0, 0.0,
0.0, size.y, 0.0, 0.0,
0.0, 0.0, size.z, 0.0,
0.0, 0.0, 0.0, 1.0);
mat4 model = translate * scale * quaternion;
vec3 n = normalize(cameraPosition - lookAtPosition);
vec3 u = normalize(cross(upVector, n));
vec3 v = cross(n, u);
mat4 view=mat4(u.x,v.x,n.x,0,
u.y,v.y,n.y,0,
u.z,v.z,n.z,0,
dot(-u,cameraPosition),dot(-v,cameraPosition),dot(-n,cameraPosition),1);
mat4 modelView = view * model;
float p11=((2.0*near)/(right-left));
float p31=((right+left)/(right-left));
float p22=((2.0*near)/(top-bottom));
float p32=((top+bottom)/(top-bottom));
float p33=-((far+near)/(far-near));
float p43=-((2.0*far*near)/(far-near));
mat4 projection = mat4(p11, 0, 0, 0,
0, p22, 0, 0,
p31, p32, p33, -1,
0, 0, p43, 0);
//lighting calculation
vec4 vertexInEye = modelView * vec4(vPosition, 1.0);
vec4 lightInEye = view * lightPosition;
vec4 normalInEye = normalize(modelView * vec4(vNormal, 0.0));
lPos = lightInEye.xyz;
vPos = vertexInEye.xyz;
vNorm = normalInEye.xyz;
gl_Position = projection * modelView * vec4(vPosition, 1.0);
}
虽然算法变形目标动画作品，我得到缺少的最终计算混合变形的面孔。动画有点貌似后续GIF。
Although the algorithm for morph target animation works, I get missing faces on the final calculated blend shape. The animation somewhat looks like the follow gif.
该blendshapes是由被称为FaceShift一个无标记的面部动画软件出口。
The blendshapes are exported from a markerless facial animation software known as FaceShift.
但同时，该算法完美的作品与它正常的长方体是在搅拌机中创建扭曲的混合变形：
But also, the algorithm works perfectly on a normal cuboid with it's twisted blend shape created in Blender:
莫非什么毛病我使用的面部动画的blendshapes？还是我做错了什么在顶点着色器？
Could it something wrong with the blendshapes I am using for the facial animation? Or I am doing something wrong in the vertex shader?
--------------------------------------------------------------Update----------------------------------------------------------
因此，作为建议，我就需要顶点着色器的变化，并提出了新的动画，还是我得到了相同的结果。
So as suggested, I made the changes required to the vertex shader, and made a new animation, and still I am getting the same results.
下面是更新后的顶点着色器code：
Here's the updated vertex shader code:
#version 400 core
in vec3 vNeutral;
in vec3 vSmile_L;
in vec3 nNeutral;
in vec3 nSmile_L;
in vec3 vSmile_R;
in vec3 nSmile_R;
uniform float left;
uniform float right;
uniform float top;
uniform float bottom;
uniform float near;
uniform float far;
uniform vec3 cameraPosition;
uniform vec3 lookAtPosition;
uniform vec3 upVector;
uniform vec4 lightPosition;
out vec3 lPos;
out vec3 vPos;
out vec3 vNorm;
uniform vec3 pos;
uniform vec3 size;
uniform mat4 quaternion;
uniform float smile_w;
void main(){
float neutral_w = 1.0 - smile_w;
float neutral_f = clamp(neutral_w, 0.0, 1.0);
vec3 vPosition = neutral_f * vNeutral + smile_w/2 * vSmile_L + smile_w/2 * vSmile_R;
vec3 vNormal = neutral_f * nNeutral + smile_w/2 * nSmile_L + smile_w/2 * nSmile_R;
mat4 translate = mat4(1.0, 0.0, 0.0, 0.0,
0.0, 1.0, 0.0, 0.0,
0.0, 0.0, 1.0, 0.0,
pos.x, pos.y, pos.z, 1.0);
mat4 scale = mat4(size.x, 0.0, 0.0, 0.0,
0.0, size.y, 0.0, 0.0,
0.0, 0.0, size.z, 0.0,
0.0, 0.0, 0.0, 1.0);
mat4 model = translate * scale * quaternion;
vec3 n = normalize(cameraPosition - lookAtPosition);
vec3 u = normalize(cross(upVector, n));
vec3 v = cross(n, u);
mat4 view=mat4(u.x,v.x,n.x,0,
u.y,v.y,n.y,0,
u.z,v.z,n.z,0,
dot(-u,cameraPosition),dot(-v,cameraPosition),dot(-n,cameraPosition),1);
mat4 modelView = view * model;
float p11=((2.0*near)/(right-left));
float p31=((right+left)/(right-left));
float p22=((2.0*near)/(top-bottom));
float p32=((top+bottom)/(top-bottom));
float p33=-((far+near)/(far-near));
float p43=-((2.0*far*near)/(far-near));
mat4 projection = mat4(p11, 0, 0, 0,
0, p22, 0, 0,
p31, p32, p33, -1,
0, 0, p43, 0);
//lighting calculation
vec4 vertexInEye = modelView * vec4(vPosition, 1.0);
vec4 lightInEye = view * lightPosition;
vec4 normalInEye = normalize(modelView * vec4(vNormal, 0.0));
lPos = lightInEye.xyz;
vPos = vertexInEye.xyz;
vNorm = normalInEye.xyz;
gl_Position = projection * modelView * vec4(vPosition, 1.0);
}
另外，我的片段着色器看起来是这样的。 （相对于前面我刚添加了新的材质设置）
Also, my fragment shader looks something like this. (I just added new material settings as compared to earlier)
#version 400 core
uniform vec4 lightColor;
uniform vec4 diffuseColor;
in vec3 lPos;
in vec3 vPos;
in vec3 vNorm;
void main(){
//copper like material light settings
vec4 ambient = vec4(0.19125, 0.0735, 0.0225, 1.0);
vec4 diff = vec4(0.7038, 0.27048, 0.0828, 1.0);
vec4 spec = vec4(0.256777, 0.137622, 0.086014, 1.0);
vec3 L = normalize (lPos - vPos);
vec3 N = normalize (vNorm);
vec3 Emissive = normalize(-vPos);
vec3 R = reflect(-L, N);
float dotProd = max(dot(R, Emissive), 0.0);
vec4 specColor = lightColor*spec*pow(dotProd,0.1 * 128);
vec4 diffuse = lightColor * diff * (dot(N, L));
gl_FragColor = ambient + diffuse + specColor;
}
最后，我从更新code拿到了动画：
And finally the animation I got from updating the code:
正如你所看到的，我仍然得到一些失踪的三角形/面的变形目标动画。关于该问题的任何更多的建议/意见将是非常有益的。再次感谢提前。 :)
As you can see, I am still getting some missing triangles/faces in the morph target animation. Any more suggestions/comments regarding the issue would be really helpful. Thanks again in advance. :)
更新：
因此，作为建议，我翻转法线，如果点（vSmile_R，nSmile_R）LT; 0 ，我得到了下面的图像效果。
So as suggested, I flipped the normals if dot(vSmile_R, nSmile_R) < 0 and I got the following image result.
此外，而不是从OBJ文件获取法线，我试图计算我自己（的脸和顶点法线），仍然我得到了同样的结果。
Also, instead of getting the normals from the obj files, I tried calculating my own (face and vertex normals) and still I got the same result.
推荐答案
我有<一个href=\"http://blenderartists.org/forum/showthread.php?290965-problem-exporting-skeletal-animation-to-multiple-OBJ-files\"相对=nofollow>一个非常类似的问题的前一段时间。当你最终发现，你的问题很可能出在网本身。就我而言，这是不一致的网状三角。使用解决的问题为了我。也许你应该尝试一下了。
I had a very similar problem some time ago. As you eventually noticed, your problem most probably lies in the mesh itself. In my case, it was inconsistent mesh triangulation. Using the Triangluate Modifier in Blender solved the problem for me. Perhaps you should give it a try too.
这篇关于出问题面部变形目标动画在OpenGL（C ++）的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持！