I'm interested in wikitude from the casual gaming perspective. There are scenarios where we'd have to naviagate the character to walk around the space, but it would have to fall from a desk to the floor. Or it would have to climb from street to pavement.
I have only done simple AR with markers in the past. For the scenarios described above I can imagine image recognition would have to do edge detection, segmentation, as well as determining X,Y,Z, angles, and curvatures of those segments. Hardly impossible IMHO, unless Wikitude has magic powers.
If this is not possible, then what ekse *IS* possible in lcoal 3D image recognition, so that we can tweak the game plays around that.
about 5 years ago
Our current 3D Tracking beta version is optimised for small scenes. You can find details on the latest news and current status here. Best is to try it out for your use case with our Natvie API or our Unity Plugin.
The best way to start is to check out the sample app which comes with the SDK download package together with the respective documentation section.
Each section includes a set-up guide, which helps you set up your project and includes a 3D tracking sample.