We have been approached by a UNESCO temple to think about how to provide an AR experience for the visitors. There are both outdoors and indoors points of interests e.g. roof top artefacts (outdoor GPS + Image recognition? + Compass + gyroscope) and indoor artefacts (no indoor GPS, no wifi triangulation so only can use Compass + gyroscope + image recognition?).
I have mocked up 2 examples in these screenshots.
There are over 50 POIs they want user to be able to "find out more" and some are interlinked e.g. user could choose a "path" and see POI A then POI B in sequence to find the story behind why it was done back then etc.
Our problem now is:
1. Indoor: How do we do recognition of the artefacts? It's difficult to place markers everywhere and some are high up, so user on ground can't scan that high for wikitude to recognise.
Looking at roof carvings (indoor looking up) https://www.dropbox.com/s/lpjqpybelr1u7yh/Screenshot%202018-09-29%2018.01.54.png?dl=0
Interior (door paintings, wall carvings etc) https://www.dropbox.com/s/96eoso9w4quovjr/Screenshot%202018-09-29%2018.03.20.png?dl=0
2. Outdoor: We can use a combination of geo tags but is it possible to exactly position a POI button on the exact position.
Across the road looking at temple roof https://www.dropbox.com/s/pwwfnxzx57lzbkc/Screenshot%202018-09-29%2018.00.40.png?dl=0
3. How will new wkitude 3D object mapping work to our advantage, I've seen scanning AROUND a 3D object but not being inside an object and scanning the environment e.g. user inside temple looking around.
Our current idea is doing a 360 shoot and making the experience VR instead - with hotspots in the VR which corresponds to users' direction + gyro and they can visually map the hotspots to what they are seeing in the temple itself. Not sure how to do this in AR instead which will be much cooler.
How is recording done? I've done one object recognition where we recorded a small sculpture and removed the environment in video to isolate the object. But this is from user looking at a smaller object. https://www.wikitude.com/blog-object-recognition-multi-target-tracking-hit-testing-slam-track-world-sdk-7/
In our case is user being "inside" the space and looking around.
I'm looking at this video, how are you doing this?
1. instant target - this is from a point of view, looking around environmental and placing markers to be viewed later. If this is possible then I can go around the space to "place markers" so users can then view these points of interest.
2. extended object recognition - again, this is also some pre-recorded marker (in this case the statue) being recognised. It is also from point of view of someone (smaller scale) looking towards something around (larger scale) the environment (and not of a larger scale person looking around a smaller object) - pivoting out towards, instead of pivoting in towards.
Starting with version SDK 8.0, Object Taregts can be generated from a set of images instead of a single video. For more information on how to work with that, please refer here https://www.wikitude.com/external/doc/documentation/latest/android/targetmanagement.html#object-targets.
To your questions:
1. Did you try to work with our object recognition - this might solve the problem of the recognition and you don't need to put targets on it.
2. Sounds feasible - pls note that the accuracy of the displayed POIs is highly dependent on the accuracy of the sensor values that the SDK receives from the system.
3. The object reco quality highly depends on the object that you wish to recognize. You'll need to test the objects to see if they are suitable.
Thx and greetings