I am trying to make 3DR texturing but it always use only vertex colors in texture.
On every frame I store frame as PNG:
RGBImage frame(t3dr_image, 4);
std::ostringstream ss;
ss << dataset_.c_str();
ss << "/";
ss << poses_.size();
ss << ".png";
frame.Write(ss.str().c_str());
poses_.push_back(t3dr_image_pose);
timestamps_.push_back(t3dr_image.timestamp);
In the method Save I am trying to process texturing:
1) I extract full mesh from context
Tango3DR_Mesh* mesh = 0;
Tango3DR_Status ret;
ret = Tango3DR_extractFullMesh(t3dr_context_, &mesh);
if (ret != TANGO_3DR_SUCCESS)
std::exit(EXIT_SUCCESS);
2) Create texturing context using extracted mesh
Tango3DR_ConfigH textureConfig;
textureConfig = Tango3DR_Config_create(TANGO_3DR_CONFIG_TEXTURING);
ret = Tango3DR_Config_setDouble(textureConfig, "min_resolution", 0.01);
if (ret != TANGO_3DR_SUCCESS)
std::exit(EXIT_SUCCESS);
Tango3DR_TexturingContext context;
context = Tango3DR_createTexturingContext(textureConfig, dataset.c_str(), mesh);
if (context == nullptr)
std::exit(EXIT_SUCCESS);
Tango3DR_Config_destroy(textureConfig);
3) Call Tango3DR_updateTexture with data I stored before (this does not work)
for (unsigned int i = 0; i < poses_.size(); i++) {
std::ostringstream ss;
ss << dataset_.c_str();
ss << "/";
ss << i;
ss << ".png";
RGBImage frame(ss.str());
Tango3DR_ImageBuffer image;
image.width = frame.GetWidth();
image.height = frame.GetHeight();
image.stride = frame.GetWidth() * 3;
image.timestamp = timestamps_[i];
//data are for sure in this format
image.format = TANGO_3DR_HAL_PIXEL_FORMAT_RGB_888;
image.data = frame.GetData();
ret = Tango3DR_updateTexture(context, &image, &poses_[i]);
if (ret != TANGO_3DR_SUCCESS)
std::exit(EXIT_SUCCESS);
}
4) Texturize mesh
ret = Tango3DR_Mesh_destroy(mesh);
if (ret != TANGO_3DR_SUCCESS)
std::exit(EXIT_SUCCESS);
mesh = 0;
ret = Tango3DR_getTexturedMesh(context, &mesh);
if (ret != TANGO_3DR_SUCCESS)
std::exit(EXIT_SUCCESS);
5) Save it as OBJ (in the result texture are only data from vertex colors, why?)
ret = Tango3DR_Mesh_saveToObj(mesh, filename.c_str());
if (ret != TANGO_3DR_SUCCESS)
std::exit(EXIT_SUCCESS);
ret = Tango3DR_destroyTexturingContext(context);
if (ret != TANGO_3DR_SUCCESS)
std::exit(EXIT_SUCCESS);
ret = Tango3DR_Mesh_destroy(mesh);
if (ret != TANGO_3DR_SUCCESS)
std::exit(EXIT_SUCCESS);
All methods returned TANGO_3DR_SUCCESS.
Full code here: https://github.com/lvonasek/tango
Thanks for reaching out and providing the detailed code breakdown.
The error is on our end - the library currently doesn't support RGB texture inputs. It assumes YUV for all input images. I've opened a ticket to track this bug and we'll fix it for the next release, by allowing RGB input and providing better return values for invalid image formats.
Edit: Found another bug on our end. The API states image_pose should be the pose of the image, but our implementation actually expects the pose of the device. I've opened a bug, and this will be fixed in next release (release-H).
You can try working around this for now by passing in the device pose without multiplying the device-to-camera extrinsic calibration, although of course that's just a temp bandaid.