3D Scanning with Scaniverse

The first object I scanned was fairly large, and I was surprised by how easily the scanning software was able to capture it. I expected the size to make the process more difficult, but the scan came together smoothly. One of the most interesting aspects was how the software extrapolated features that I could not physically capture. For example, I was not able to raise my arm high enough to take images of the top of the trailer, yet the software was still able to reasonably infer and reconstruct what the top looked like. I was suprised to see how much noise remained on completely flat surfaces like the large walls of the trailer but I can see this happening with the reflection of the metal. The result was more accurate than I anticipated, which highlighted how powerful the underlying algorithms can be even with incomplete data.

The second object I scanned was the Wines sign. Similar to the first scan, I found it interesting how the software handled areas that were not fully captured. In particular, the tops of the posts on either end of the sign were extrapolated in a way that still looked realistic and consistent with the rest of the model. This again demonstrated the software’s ability to make logical assumptions based on the surrounding geometry. Seeing this repeated across different objects increased my confidence in how well 3D scanning can handle missing or hard-to-reach perspectives.

The final object I scanned was a playground, which I expected to be the most challenging due to its complex structure. Because it is not a solid object and contains many holes, bars, and pass-through areas, I assumed the software would struggle to reconstruct it accurately. However, it performed surprisingly well and was able to capture a high level of geometric detail. One drawback I noticed across the scans was the amount of noise present in the final models. The surfaces appeared rough, which makes sense given the amount of information being processed in each image. This roughness would make it difficult to export the models into formats like STL without additional cleanup.

This technology is extremely impressive and exciting to see in practice. When I first learned about 3D scanning in class, I initially thought about its potential use for modeling in the context of 3D printing, since that is something I am interested in. However, it quickly became clear that scanning objects using a phone-based photogrammetry approach would not produce models suitable for 3D printing. This makes sense, as models intended for manufacturing need to be highly precise and free of discontinuities. While photogrammetry creates visually impressive models with a high level of detail, they often lack the precision and smooth continuity required for physical production.

That said, this technology would be very effective for bringing real-world objects into digital environments such as video games or 3D rendering engines like Blender. Keeping the models in a purely digital space avoids many of the limitations associated with physical fabrication and allows the technology to be used in a way that better aligns with its strengths.