Usage
The Foto3D CX software runs on both Mac and Windows, and for my tests I ran it on a lightweight MacBook (pre-aluminum unibody version). I used a 10MegaPixel Canon digital camera with built-in flash for the photos, and nothing fancy for lighting or backgrounds (which I’ll get into more about later).
The first step is to simply photograph your model. For my first test I used one of my daughter’s small plastic bear toys (shown below), placed directly on the marker.
Test Subject #1
The CS graduates among you will recognize that the pattern is a circular representation of binary digits scattered in a circle. Also, you’ll see here that my lighting was far from ideal. Strata recommends that you use a background of high-contrast from your model, in their demonstrations they use a blown-out all-white background. They also recommend a short pedestal to lift the model an inch or two above the marker.
I proceeded to take 18 pictures: 16 around the model, and two from higher angles to better capture the top of the head. Those images were then loaded into Foto3D and I told it to go ahead and begin the “Fully automatic” process. In fully automatic mode it automatically segments the model from the background, automatically registered the images to each other, automatically generates the texturemap and refines the geometry. Unfortunately, the automatic process was unable to fully segment the bear in a few of the pictures requiring me to manually do it.
Manually performing the segmentation is easy to do, tho. You can tweak the parameters on the automatic segmentation to get close, then use a simple Paintbrush & Eraser to adjust the areas of the “Mask”. After manually tweaking some of the images (as you see above), I told it to generate the geometry again, and it worked.
Here you can see a wireframe version of the resulting geometry. The tessellation is done “nicely”, with no degenerate or skinny triangles. However, you can see that the geometry is only a rough approximation of the actual shape of the bear. The arms and legs are visible, but the details of the face simply aren’t there, because they are fairly minor to begin with.
The next step is to tell it to generate the texture, which analyzes the photos and geometry again to generate color information.
The final textured 3D Bear.
Here you see the final result. The still image doesn’t really do it justice, but the result is actually very well-done for the minimum of work I put into it. Masking the photographs took about 10 minutes, and my computer was capable of generating the 3D Model and Texture Map in under 10 seconds each. The color information does a good job of hiding the lack of detail in the 3D model, and really “completes” the model.
With this complete, you have a few options:
- You can export the model and textures as a WRL object, that you can then import in various packages for further manipulation.
- You can use the Foto3D CX Plugin for Photoshop to import this into a PSD file as a 3D object. Once the object is in Photoshop you can rotate & translate it in 3D, and scale it to fit into your image.
Having gotten the hang of the tool with the bear, I decided to try something more complicated: The man of Steel himself.
This software is complete waste of money. After spending most of two days trying to coax it to work. I give up. It just doesn’t work. Neat idea. Bad implementation.
don’t waste your time or money for this crap. Real 3d artists never use it.
I like the article on the 3d scanner, they have come a long way in recent years. I personally use a 3d scanner from 3d digital corp that is similar in function to the cyberware scanner.