Best Buy
  • Price: £500 inc VAT

  • Company: D Vision Works

  • Our Rating: We rate this 9 out of 10 We rate this 9 out of 10

D Sculptor extracts textured 3D models from conventional photographs in a simpler way than rival systems that use specialized scanners or cameras. It’s mainly suited to modelling small- to medium-sized discrete objects, though in theory any size can be accommodated.

D Sculptor extracts textured 3D models from conventional photographs in a simpler way than rival systems that use specialized scanners or cameras. It’s mainly suited to modelling small- to medium-sized discrete objects, though in theory any size can be accommodated.

D Sculptor works by comparing a series of images taken from different angles, and extracting a polygonal mesh. The photographic images are also converted to a texture map that’s applied to the mesh surface. The result is a 3D model with a true photographic surface that conveys all the tonal subtleties of the original object without having to create material, bump and reflection maps. Although the polygon mesh models aren’t always totally accurate, the texture map hides small irregularities and the end result can be excellent

The drawback is that shadows and specular highlights in the original pictures are faithfully reproduced, limiting your ability to place models into new environments and match the lighting. I found it best to photograph objects in flat, nearly shadowless lighting – for glossy surfaces, the developer recommends shooting with polarizing filters over flash and lens.

The object has to be placed on a square of target points for photography. These are provided on-disk as printable files in several formats. In principle, you arrange separate targets around a large object such as a car, moving the camera around it.

Some types of original won’t work, such as transparencies (glass or plastic objects), or objects with dark hollows. Where under-surfaces are important (such as aircraft wings), objects can be suspended or supported at an angle so the camera sees them.

With the pictures taken, you import them as a group into D Sculptor. You then click on the centres of all visible targets to number them in each picture. Next you have to create a mask around the object’s outline – the more you mask, the better the accuracy.

The mask tools are manual and basic compared to an image editor like Photoshop, so it’s a pity there’s no facility to import an alpha channel. There’s a simple rectangle fill tool, plus a manual polygon creator that most models will need. There’s a basic snap-to-edges feature, but I found this inaccurate, and had to create a lot of manual points. Holes in objects can be created by switching to a background mask creator.

When you’ve created enough masks, switch to the model view and click to extract the mesh. There’s a control for the number of polygons and voxel density for detail. This can take several minutes depending on the speed of your PC and the number of pictures. The resulting model is displayed as a wireframe or simple shaded surfaces, and can be rotated onscreen. Next, you click to apply the texture map. This usually looks good first time, but there may be mismatches due to conflicts between the photographs. You can go back into the original photographs and force D Sculptor to include or ignore small areas of texture.

Once complete, a model can be exported in VRML, Wavefront and Direct-X formats with texture maps, or DXF without. You can export 2D bitmap snapshots, with an option of inserting a background image. There’s a useful Java export option that also generates HTML code so you can paste it straight into your Web site.

For real-world models for Web sites or games, D Sculptor is an easy, fairly fast and relatively inexpensive alternative to CAD-oriented laser scanners and cameras.