Hi
I have been working with RC since an year now and have been able to successfully simplify the meshes and use it in Unreal Engine. However, since I now plan to 3D print the models, I am finding it a major problem to get the details of the model for the print.
I have checked various videos on the subject with very limited success. Following methods have been used by me so far.
Texture reprojection - to get Displacement map and use it in blender to displace the mesh.
The map doesn’t seem to be capturing the depth data and the visual representation shows only blue and green shades with no red shades.
Convert normal/diffuse map to Height Map using Materialize: This process also did very little to convert the details and change the mesh.
I am enclosing the screenshot of the render in Reality Capture (RC_Image), screenshot of the model in Blender (Blender_Image) and the screenshot of the displacement map in Reality Capture.
Would appreciate any help on this as I would like to print these with the details.
PS: I have already started sculpting in blender to clean up and highlight the details. But this is a very laborious process and the detailing is not very accurate (Sculpt_Image).
I am curious. Why are you using the displacement map in Blender to displace the geometry for 3D printing instead of the original high-poly mesh?
Another question is whether the 3D printer will be able to print such details. Of course, that depends on the size of the print, the printer, etc.
Anyway, the colors you see in RealityCapture are just a visual representation of the data. The data has no colors, only positive and negative values from the texture reprojection tool. Positive values are above the simplified mesh surface, and negative values are below.
In order to use this displacement map in 3D engines, it usually has to be normalized between 0 and 1 values. It is also doable in Blender.
In order to displace the mesh, it needs to have enough geometry (polygons) that can be displaced. That might be the reason why you didn’t see much change because you were trying to apply the displacement on a low poly mesh.
For more information about the texture projection and displacement maps check this tutorial about texture projection.
This is not a low-poly mesh. Its a relatively hi-poly mesh of 15.0M tris. I simplified it from a 19.6M tris mesh just to get the normal maps.
You can see the difference between Solid and Sweet inside RC application itself. I am enclosing the screenshots.
This is the main reason I tried creating a Displacement map as there is considerable loss of detail in the Solid model.
Yes I followed the texture reprojection tutorial and that’s the reason why I mentioned that “visual representation” does not show any Red color. From what I understand, while ‘blue’ color represents the height above the surface, the ‘red’ indicates the depth. In this instance, the displacement representation ranges from light blue to yellow and orange which could indicate that the values are mid-range. And hence very little effect in blender on the mesh.
As mentioned, I have started sculpting the model (15.0M tris) in blender to enhance the details and imported it into a slicer application.
The details come out good in the areas sculpted by me vs the areas not worked on.
Hi @GPuthige,
you can’t compare textured and not textured mesh. The textures could cover the mesh, but there is only the mesh you want to use (as for the textures the images are used, you can see it as more clear, but it is still the same mesh, sweet is basically solid covered with texture, so there is no loss of detail it is still the same).
What is the used camera resolution? What were your reconstruction settings?
Thanks for the revert and the clarification on the texture vs non-textured mesh.
More info on the capture :
A total of 238 images captured this 12th Century Sculpture
The images were taken on OnePlus OnePlus 10T 5G
The image resolution is 4096 x 2304 pixels in 32bit RGBA heic format
The model was calculated in normal quality.
I am enclosing the image with Alignment setting and Color & Texture setting.
Our objective is to generate a mesh with the maximum possible details of the sculpture and as accurate as possible. Since there are very intricate details, we are finding it difficult to 3d print an accurate model.
Thanks for your help on this.
PS: This is being done by us as part of the Indian Heritage and Culture Conservation program since this sculpture is about 850 years old. Any suggestions, references to training and workshops would be most welcome.
Hi,
first of all, for such work you should use better camera. Full-frame camera bodies because of price / performance ratio (a sweet spot is around 40 Mpix because of quality and time to process the data, higher megapixel count requires a shorter shutter speed (max 250/s) or tripod) or digital medium format camera bodies for best quality. Maybe more images also with detailed one.
Also, you can use High reconstruction.
But following all of these and then printing 3D model you can lose the details anyway.
