FAQ – scAnt3d

scAnt combines automated multi-angle image capture, image stacking and masking, and Structure-from-Motion / multi-view stereo photogrammetry to produce high-quality textured meshes (Have a look at the original publication for more info scAnt paper, Plum & Labonte, 2021). In the future we aim to support modern 3D reconstruction options such as Gaussian Splats.

It primarily uses photogrammetry.

No — scAnt is designed to produce high quality scans “on-the-fly” so where required, calibration is largely automated. We provide extensive documentation for scanning parameter choices and presets for common cameras.

“Ant to Apple”; scAnt’s default configuration supports capturing subjects between 5 and 50 mm. For very small high-magnification requirements, lenses and extension tubes can be interchanged.

Scan accuracy depends on camera resolution, optics, working distance and scanned object. While the minimum resolvable detail size depends on the object’s dimensions and visual properties, we have captured geometric details at high accuracy of up to 0.05 mm. Higher resolution in colour textures is possible still and dependent on the choice of optics.

Shiny or transparent surfaces are challenging for photogrammetry; scAnt can be equipped with a cross-polarised lighting system to reduce the reflectivity of objects. When no colour information is needed, consider coating your subject with 3D scanning spray.

scAnt supports several camera classes: mirrorless/DSLR, and industrial machine-vision modules; different bundles will list the included camera and lens options.

Depending on the object size and capture step settings, a typical scan takes between 60 to 120 minutes.

None! scAnt has its own dedicated lighting solution built in.

As long as the object is rigid and does not deform while being moved, no specific preparation is needed, other than that it needs to be attached to the object holder. There are different attachment options available depending on what type of object you want to scan.

scAnt Studio allows you to setup scans, run calibration and tests, and process captured images. Future versions will add advanced processing options, combining open-source SfM/depth engines and custom fusion/texture tools; we’ll support export to common formats for photogrammetry and gaussian splatting workflows. (Exact feature list & licensing details to be announced.)

Yes — scAnt outputs standard image sets and camera metadata usable in tools like Meshroom, Zephyr, Metashape, or RealityCapture. We mostly use Meshroom and Zephyr in our current work.

For smooth, high-quality dense reconstruction we recommend a modern multi-core CPU, 16 GB+ RAM, and an NVIDIA CUDA GPU. As the raw images of scAnt scans are often in the thousands and may be stored in loss-less compression form, they can occupy up to 100 GB of storage for a single scan. For most systems the bottleneck thus becomes hard drive capacity and writing speed. We recommend using modern SSDs with at least 2 TB of storage for high throughput scanning.

We plan to offer both local and optional cloud processing (pay-per-use or subscription) for users without a high-end GPU. (Cloud timing & pricing TBA.)

Basic cleanup (remove floating noise, hole fill, decimation) is optional — we provide ample documentation on how post processing can be done with open source software such as blender; advanced retopology or CAD repair is supported through standard toolchains.

Yes — scAnt photogrammetry pipelines outputs meshes that can be simplified, watertight-repaired, and exported for 3D printing or imported into CAD packages after conversion.

scAnt captures high-quality textures; for critical colour work we recommend colour calibration targets. We are working on an automated routine for colour calibration that will be available with future releases of scAnt Studio.

FAQs

The Team