3D scanning isn’t nearly as common as 3D printing, but the technology has still improved dramatically in recent years. The goal, at least for makers, is to be able to 3D scan an object with enough detail and accuracy to be able to then 3D print an exact duplicate of that object. There are different methods for 3D scanning, but the two most common are laser scanning and photo scanning. Laser scanning is generally better for creating dimensionally-accurate 3D models, but it still struggles to pick up the fine detail on small models. The new, open source scAnt ditches the laser and picks up fine detail on small objects using photo scanning.
Photo scanning, or photogrammetry, works by taking hundreds of photos of an object from every angle. Those photos are then stitched together to form a 3D model. In order for photogrammetry to work properly, the photos all need to be taken from a consistent distance, with identical focal length, and with equal illumination. While simple 3D models can be created from videos or photos taken from a handheld camera, the results are very poor. To get the best results, desktop 3D scanners are generally setup as special rigs that snap the photos automatically. That is true for scAnt, but it has also been optimized for scanning really small things, like insects.
The object to be scanned is held within a spherical housing lined with LEDs to provide the proper illumination. The camera points through a small opening in that sphere. That objects inside is attached to a special mount that rotates in two axes, which makes it possible to take photos from every angle. Because scAnt is open source, you can build it yourself. You just need the 3D printed and laser-cut parts, stepper motors, Pololu USB stepper drivers, a FLIR Blackfly camera, and a computer. That camera is fairly expensive at £289.00 (about $400), but other cameras can be substituted — the results just may not be as good.
And the results from scAnt are very good. It is capable of producing nearly photorealistic 3D models, which opens up some unique possibilities. Specifically, the developers of scAnt plan to use the scanned 3D models of insects to create synthetic data sets for training machine learning models. The insect 3D models can be posed automatically and images from every angle can be rendered, which means that thousands of training images can be created without any manual labor. Those data sets can then be used to train machine learning models to recognize the scanned insects in the wild.