Frames received by client only contain a row-vector of pixels. about syphon-framework

[bangnoise]

Due to how Apple's IOSurface/OpenGL intergration works, calls to glGetTexImage() won't work - if you need access to the pixel data, draw the texture to an FBO and then download from the FBO's colour attachment (ideally into a PBO so you can defer the download).

Obviously drawing to the window via this method is a terrible idea (just draw the original texture directly) - but presumably this is just an example and you plan to do further processing via OpenCV.

[Kautenja]

Ah, I see. Thanks for the feedback. For future reference, rendering to a frame buffer did address the issue. This was the client callback block that solved my issue based on your suggestion. Indeed the goal is to process the frame, rendering with OpenCV is just a sanity check.

                    SyphonOpenGLImage *frame = [client newFrameImage];

                    GLuint fbo;
                    glGenFramebuffers(1, &fbo);
                    glBindFramebuffer(GL_FRAMEBUFFER, fbo);
                    glFramebufferTexture2D(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_RECTANGLE, frame.textureName, 0);
                    GLenum status = glCheckFramebufferStatus(GL_FRAMEBUFFER);
                    if (status != GL_FRAMEBUFFER_COMPLETE) {
                        std::cout << "frame buffer error " << status << std::endl;
                    }

                    glBindTexture(GL_TEXTURE_RECTANGLE, frame.textureName);
                    int width, height;
                    glGetTexLevelParameteriv(GL_TEXTURE_RECTANGLE, 0, GL_TEXTURE_WIDTH, &width);
                    glGetTexLevelParameteriv(GL_TEXTURE_RECTANGLE, 0, GL_TEXTURE_HEIGHT, &height);

                    uint8_t* pixelBuffer = new uint8_t[width * height * 4];
                    std::memset(pixelBuffer, 0, width * height * 4);

                    glViewport(0, 0, width, height);
                    glClearColor(0.0f, 0.0f, 0.0f, 0.0f);
                    glClear(GL_COLOR_BUFFER_BIT);

                    glEnable(GL_TEXTURE_RECTANGLE);
                    glDisable(GL_DEPTH_TEST);

                    glBindTexture(GL_TEXTURE_RECTANGLE, frame.textureName);

                    glBegin(GL_QUADS);
                    glTexCoord2f(0.0f, 0.0f); glVertex2f(-1.0f, -1.0f);
                    glTexCoord2f(width, 0.0f); glVertex2f(1.0f, -1.0f);
                    glTexCoord2f(width, height); glVertex2f(1.0f, 1.0f);
                    glTexCoord2f(0.0f, height); glVertex2f(-1.0f, 1.0f);
                    glEnd();

                    glReadPixels(0, 0, width, height, GL_RGBA, GL_UNSIGNED_BYTE, pixelBuffer);

                    glBindTexture(GL_TEXTURE_RECTANGLE, 0);
                    glDeleteFramebuffers(1, &fbo);

                    torch::Tensor tensor = torch::from_blob(pixelBuffer, {height, width, 4}, torch::kUInt8).flip(0).slice(2, 0, 3, 1).contiguous();
                    cv::imshow(serverName, tensor2mat(tensor));
                    char c = (char) cv::waitKey(10);

                    delete[] pixelBuffer;
                    [frame release];