always return cv:Mat()

[rootkit]

when the app run here:
cv::Mat FaceSwap::blend(const cv::Mat& src, const cv::Mat& dst, const cv::Mat& dst_seg)
if (minc >= maxc || minr >= maxr) return cv::Mat();

blend always return cv:Mat(), I don't know what happen?
BTW: I finish port the great app to iOS platform, in the end run here and stop, return failstd::runtime_error: Face swap failed!

Thanks

John

[YuvalNirkin]

This means that the combined segmentation map is only zeros.
It might be a problem with the landmarks or the segmentation.

I like the idea of an iOS port! It can be a great contribution! Are you willing to share the code?

[rootkit]

the log:
faces found = 1
faces found = 1
vecR [0.0029098843;
0.15772882;
-0.037092187]
vecT [1.9298197;
-21.032534;
-1098.7878]
exprW [-0.39331913;
0.17700461;
-0.29514793;
0.11593564;
-0.42752558;
0.24469842;
2.8671944;
0.7221989;
-1.1991261;
0.099073336;
0.20576096;
1.0124749;
-0.44310626;
-0.098507486;
-1.0734005;
-0.24701883;
-0.37821278;
0.3187601;
1.1587392;
1.8510281;
1.3205831;
0.76069188;
-0.029121373;
-0.0070772283;
1.1604365;
0.0073522874;
0.24649803;
-0.02893436;
-0.24754204]
vecR [-0.034498993;
0.077174194;
-0.13330679]
vecT [4.0268807;
-32.977039;
-924.23651]
exprW [-0.24939266;
0.39104307;
-0.17418325;
0.77485698;
-0.11051087;
0.25302994;
1.5111284;
-1.7924082;
-0.21672148;
-0.19027816;
-0.27793592;
1.3636755;
0.03324154;
-0.012820265;
-1.6727974;
-0.12949848;
-0.047394246;
0.91332167;
0.16732536;
0.04252547;
0.11279398;
0.49106297;
-0.008773597;
-0.21201798;
1.1952631;
0.24981749;
0.021493442;
-0.29385114;
-0.26967478]
libc++abi.dylib: terminating with uncaught exception of type std::runtime_error: Face swap failed!

I found here :
cv::Mat rendered_img;
m_face_renderer->render(m_vecR, m_vecT);
m_face_renderer->getFrameBuffer(rendered_img);
rendered_img data is zeros, I think the issues is here? Am I right?

I will share the porting code till the app can run on iOS, but I might not enough time to support the code.

[YuvalNirkin]

I can't track the problem from this log. What images have you used?

[rootkit]

I import some debug images, as follows:
cv::Mat debug_src_mesh_img = fs.debugSourceMesh();
cv::Mat debug_tgt_mesh_img = fs.debugTargetMesh();

cv::Mat debug_src_lms_img = fs.debugSourceLandmarks();
cv::Mat debug_tgt_lms_img = fs.debugTargetLandmarks();

cv::Mat debug_src_mesh_wire_img = fs.debugSourceMeshWireframe();
cv::Mat debug_tgt_mesh_wire_img = fs.debugTargetMeshWireframe();

BTW: face segment need to big memory more then 1GB, I have to generate them by Mac OS APP

[rootkit]

I found the rendered_img.data is zero, so opengles porting is not right?
m_face_renderer->render(m_vecR, m_vecT);
m_face_renderer->getFrameBuffer(rendered_img);

[YuvalNirkin]

What is the OpenGL ES version? It has to be equivalent to OpenGL 1.5 (without shaders).

[rootkit]

#include <OpenGLES/ES1/gl.h>
#include <OpenGLES/ES1/glext.h>
#include <OpenGLES/ES2/gl.h>
#include <OpenGLES/ES2/glext.h>

[rootkit]

I have to use opengl es1, delete es2.0, now I test it... god bless me

[rootkit]

I already change opengl es2 to es1, no lucky:), rendered_img is still zero.
I ongly add a function beacause es2 does not provide it.
void gluPerspective(double fovy , double aspect , double zNear , double zFar)
{
// Start in projection mode.
//glMatrixMode(GL_PROJECTION);
//glLoadIdentity();
double xmin, xmax, ymin, ymax;
ymax = zNear * tan(fovy * M_PI / 360.0);
ymin = -ymax;
xmin = ymin * aspect;
xmax = ymax * aspect;
glFrustumf(xmin, xmax, ymin, ymax, zNear, zFar);
}

[rootkit]

Everything is ok but rendered_img:(

[rootkit]

So how to trace the problem in the next step? do you need more screetshots about it.