In this episode of the basic turntable photogrammetry series, we will see how to reconstruct the high poly geometry from the images with transparency generated in Blender at the end of the previous episode.
Video Transcript
Hello everyone! In this episode of the basic turntable photogrammetry series, we will see how to reconstruct the high poly geometry from the images with transparency generated in Blender at the end of the previous episode.
Among other things, we will see how to interrupt the process halfway through to resize the reconstruction area and then resume from where we left off.
First, I open the Meshroom project, which I created in the second episode and saved after setting the Minimal 2D Motion field of the Feature Matching node to 2.
I import the rendered images from Blender into the Image Gallery tab. As you can see from the red shutter icon, this time the images do not contain additional information because they are no longer the original photographs but renders created in Blender.
By using the Image Masking node, which we examined in the second episode, the original data is preserved, but removing the green background will be a bit more difficult. On the other hand, it is easier to achieve better Color Keying in Blender, but at the cost of losing the original EXIF data. However, this is not a big deal since Meshroom will handle it excellently, as we will see.
I save the Meshroom project with the new data and click on an empty spot in the Graph Editor to deselect all nodes. Then, I start the reconstruction by clicking Start.
If I now take a look at the folder where I saved the project's MG file, I notice the presence of the Meshroom Cache folder and its subfolders, where Meshroom will store log files and intermediate working files for the various nodes.
This feature of Meshroom is very useful because it allows processing only specific pieces of information as they are handled by the nodes.
During processing, a progress bar at the top displays the overall progress in the pipeline, while each node also has its own progress bar. In both cases, the green color indicates that the process was successfully completed. The orange color means the process is still running. The red color indicates a failure. The cyan color shows that the node is in the execution queue.
These progress indicators can be examined in more detail in the Task Manager tab. Additional information can be found in the Log section of each node.
Another useful feature is that if I need to interrupt a process and restart it later, Meshroom will create new subfolders for the various nodes, so the previous work will not be lost, and I will still have access to the different processing steps.
Speaking of interruptions, I click Stop as soon as the Structure From Motion processing is completed, which is the first part of the pipeline, so I can review some new data and adjust the reconstruction area, known as the Bounding Box.
First of all, at the bottom of the Image Gallery tab, we can see that Meshroom has recognized and processed all 146 input images, which seems like a great result to me.
In the 3D Viewer tab, there are icons representing various cameras. These icons indicate the viewpoints from which the photos were taken. Clicking on a photo in the Image Gallery will select a corresponding camera in the 3D Viewer tab.
You can change the viewpoint in the 3D Viewer tab by left-clicking and dragging to rotate around a pivot point. Zooming is done with the mouse scroll wheel. Clicking the scroll wheel or the middle mouse button allows you to pan the view. You can reposition the pivot point by double clicking on an element within the 3D view.
Now I can select the Meshing node and enable the Custom Bounding Box option to define a custom reconstruction area. This scene is not particularly complex, so this step might seem unnecessary, but in other situations, this option can be very useful for limiting the reconstruction to only what is needed.
However, simply enabling the Custom Bounding Box option will not display the reconstruction volume. In Meshroom, selecting a node is not enough to view its specific information or features. You need to double click on it.
Now, I can define the Bounding Box interactively using the transformation manipulators that appear at its center. The arrows allow me to move the center of the Bounding Box. The squares let me resize it along one of its axes. Finally, the circles allow me to rotate it around the axes.
After defining the Bounding Box, I can click Start to resume the geometry reconstruction pipeline from where I left off.
Now, by examining a subfolder inside the project's Cache, you can see that a new folder has been created. This happened because the previous pipeline was interrupted, so the new data is being stored in a new subfolder.
These subfolders have, by default, rather generic names, but you can rename them in the Folder fields of the various nodes. This is useful if you want to run multiple tests with different parameters, as you can name the folders based on the changes made. In any case, you can simply sort the folders by date to quickly find the latest processing result.
I resumed video recording at the end of the last processing step, which is Texturing. Before starting the process, I didn't modify this node’s default parameters, so Meshroom should have created an OBJ file with a Base Color texture in EXR format, using the UDIM method for UV unwrapping.
However, in Meshroom's menu options, I don’t see anything that suggests exporting this mesh to a specific location on disk.
No worries, though, because the mesh and its texture have already been saved on disk. The file paths can be found at the bottom of the Attributes tab in the Texturing node.
More simply, the files will be located in the most recent subfolder inside Texturing.
I copy the EXR and OBJ files to the main project folder, renaming the copies as High Poly, so that I can keep the originals safe and work on the copies in Blender.
I save the changes to the project in Meshroom and close the program. I will reopen it later when I need to Bake the textures from the High Poly version to the Low Poly version. In the next episode, I will show you how to import the High Poly version into Blender and clean it up before creating the Low Poly version. See you next time!