Start a new topic

3D model recognition on Object

3D model recognition on Object



I am using wikitude to display a 3D model. I have created a target with few images which have got 2 star.

After running the application when I point the camera on to the image or pintout of the image the 3D model is displayed.

Whereas if I point the camera towards the actual object the model is not being displayed.

Please let me know if any other changes are required for this to work.

Hi there!

Can you give me more information about what you are trying to accomplish specifically in your use case?

Generally our Image Recognition engine is designed to recognize and track planar 2d targets (e.g. images) in the camera frame. 

Hi Markus,

Thanks for the reply. 

In my application I have an image of  a building. When I point the camera towards this image a 3d model of the building is displayed.

But if i point the camera towards the actual building( instead of image) the model is not shown.

So according to you this wont work is it?



Hi again!

Yes, the Image Recognition engine is specifically optimized for the recognition of planar 2d targets like images, cd, book or magazine covers. Unfortunately in your case the target would be a 3d object in the real world, which is not suitable for tracking with the engine. This is also the reason why it works with a 2d image of your 3d object.

Hope that helps,


Is there any way that we can recognize that building in real time without using the images to recognize that building?


This not possible with the current 3D tracking functionality of the SDK.

Great news for you as today we launched SDK 7, which includes your feature request, “Object Recognition”. In a nutshell, SDK 7 introduces:
  • Object Recognition
  • Multiple Image Target Recognition
  • Hit-testing API for SLAM
  • Instant tracking improvements
  • Extended recognition range
SDK 7 significantly expands Wikitude’s SLAM technology by adding object recognition and tracking capabilities to its feature set. These features enable brands and enterprises to build augmented reality experiences using a variety of real-world objects, such as toys, sculptures, architectural models, product packaging, industrial machines and more.

The workflow is simple: record a video of your object, upload it to the Wikitude Studio Manager and create your AR experience as desired. Object recognition technology creates an additional touch point to interact with users, allowing real time and 360 degrees augmented reality experiences around real world objects.


Check out this blog for a developer insight and this tutorial to see how you can create your first 'Object Target Collection'. You can download the new SDK 7 directly from our Download page. Enjoy our new cool features ;)


is it possible to detect more than one real-life buildings in Wikitude? if yes, then please refer me tutorials or link. 

How do I initialize more than one .wto file in unity under target collection

Login or Signup to post a comment