Computer Vision, 3D and the connecting factor - AI

3D search engine

by Peter
Reading Time: < 1 minute

This time a post about something we developed – a 3D search engine.

User puts in an image of an object, the engine finds in an existing repository of 3D objects the most similar objects to what’s in the photo. In this demo’s repository there are 5000 models of chairs.

How does it work?

The technology works by mapping a 2D image into a representation of the underlying 3D features of the object within the image. Similar features are extracted from the 3D objects. Then, a similarity engine searches the 3D features from the image over the 3D features from the 3D objects.

The demo works for chairs, but this is just a demo 🙂 The same principle technology could be applied to any type of object category. As long as the 3D objects have a geometric construct (mesh\stl\obj\choose your format) the engine can find it.

Future ideas

Connect this to thingiverse or grabcad or any other online 3D repository and allow for visual search over these repositories.

Leave a Comment