Start a new topic

Instant tracking with custom initializer

Hello, i'm trying to implement something using instant tracking which i'm not really sure about. Let me explain:

I have ruins, these are basically destroyed bricks basement(no walls) at an heritage site. My aim is to augment these ruins with a rebuild 3D model of ruin when it was a  building. 


What I have tried is :

1) Taking images of the ruins for Object Recognition, which did not work well as sometimes it recognized it sometimes not. The ruins happen to have dead leaves on it which i think is the cause for the bad detection. 


2)Then i switched to extended Image Recognition. The ruin actually have a stand with a printed descriptive panel fixed in cement in from of it and i have tried to make that as the Image target. Upon detection of the latter, i would then should the augmentation behind it, having an occluder for the panel in from of it. This did not work well when i scale up the 3D model size to match the real one during the scan. After the Image is lost from the view and i try to move around to see the 3D building 360 deg, the 3D augmentation disappears. 


I feel i need to get to instant tracking. But then what i want to achieve here is having the panel as initializer and having the 3D augmentation relative to that stand. I don't know if it possible. I have searched videos on youtube and what I have seen till now are users searching for a plane and manipulating 3D objects and moving them in 3D space in the scene. I don't want to allow the user to move any object or whatsoever. Nothing should be dynamic in my use case. The user will just have to scan certain objects, in my case, a descriptive stand fixed in front of the ruin and relative to that having the augmentation appearing behind.


If this is possible, can you explain me how to achieve this? I have tried Image and Object Recognition and I am new to Instant Tracking.

I am using Wikitude SDK v8.2 with Unity v2018.2.18f1 and I am testing on a phone with android Marshmallow (v6.0).


Thanks


Hi,


It depends a lot on the surrouding and the image you wish to scan - the best technique most of the time is very dependent on what actually gets scanned and needs to be displayed. This is why testing directly in front of the ruins is the best approach to say what's the best way.


Without any images and details on the scene it's hard to distinguish - from your tests, which of the 2 is workig better?


Thx and greetings

Nicola

I have tried extended Object Recognition but i'm having issue detecting the ruin i think its because of the leaves on the floor . Actually there are lots of trees there and when i took my pictures, some leaves were on the floor, then when i'm testing, no leaves are there.  - can this actually be an issue ?


Then, whenever i am trying to do some extended Object Recognition, that is, after the Object is out of view and I keep scanning, 1/2 trials, my app crashes. Is Extended O.R device restricted ? . I am using this on a SAMSUNG with android 6.0.


Thanks.

Hi,


The app should not crash if you loose the object - it might loose tracking. If the app crashes, then something is not right. Can you please debug the app and check further for log output. Also - is the standard sample app working on your test device, if you simply switch the .wto file to have your own? 


Thx and greetings

Nicola

Thanks Nicola. The sample app works and I found another workaround. But now i'm using extended Image Tracking and when i start the app, after i have detected a target, the remaining targets are not detected unless I restart the app. Is there anything that I should do like stopping the tracking before going to the next target?


Help,

Silvio

Hi,


You could e.g. integrate a button 'recognize different area', which the user clicks if they wish to scan something different. When they click it you need to stop the extended tracking - the behaviour like this is implemented in the sample app. There you also have a 'Stop extended tracking' button.


Greetings

Nicola

Thanks for your prompt response. But in my use-case, a user will just need to walk around, following a map of targets and see the augmentations. In my use case, there is no user interaction at all with the screen unless there are some kind of buttons in form of augmentation. In my case, is there a way to avoid the user to stopping it manually ? 


Thanks.

Hi,


Unfortunately the behaviour with extended tracking is that the tracking needs to be stopped before a new target an be scanned and recognized. If you work with extended tracking for your case would it be feasible to just work with one image for initalization? If the area the user scans has enough structure this might also work. As the quality is very dependant on the real environment it would be important for you test directly at the location.


Thx and greetings

Nicola

Hi,


You can find further details on Instant Tracking (or Markerless Tracking) on its intend use in our documentation - please have a look at the intro first.


In your case if you wish to recognize a pre-defined target Object Recognition or (extended) Image Recognition is the most suitable option. 


Thx and greetings

Nicola

Thanks for reply. If i use Image Recognition having the panel as Image target and add the building augmentation around it, is it a good way to do it? 

what is the best approach for my Use case plz ? 


Thanks,

Silvio

Login or Signup to post a comment