Layar Vision http://www.layar.com/
is out, and allows to use a recognition image API to create amazing apps.
This nice feature will run on their servers and will allow you 15000 hits per month without cost but
the following hits will be charged at some 10 euro cents,
however the next 10000 hits will cost you 1000 Euro monthly, which hopefully will produce some profit it they manage to sell.
I have never seen somebody selling image recognition services and somehow is a nice approach, however more testing is needed.
However, the thing that interests me the most is the
JLayar converter tool which this time reaches version 3.0 and includes some nice features like...
http://layar.pbworks.com/w/page/32587974/Release%20History%20-%203D%20Model%20Converter
- Version 3.0
What's new:
- Overview panel : added "Minimum layar version" field which indicates on which layar client version this 3D model is supported.
- Materials panel: material names are listed to ease material modification.
- Materials panel: material transparency is supported and can be manipulated using the Opacity value. The Opacity value embedded in .obj/.mtl files will be prefilled if it is present. By default, the value is 1.
- Materials panel: color blending (alpha) is supported. when blending is enabled, it means that color value of this material is blended with the color of any other objects that are rendered behind it. If it is disabled, it means the objects that are rendered behind it are hidden from view.
- Materials panel: Added "Shader" with two options - default & nolighting. The "nolighting" option is for people who want to bake lighting effects into textures instead of relying on the standard lighting from the AR view.
- "Save for version" option makes it possible to save the same model for different layar version. This is useful when you want to have your models work on various versions of layar clients. You can determine which 3d model should be returned in getPOIs response based on the "version" parameter in getPOIs request. Please NOTE that if you save the model for a specific older version, some features will be stripped from the model. For example, transparent materials might become opaque, and animated textures might be converted to a static version.
Known issues:
- Need to throw error messages when there is no material attached an .obj file and there is no material name in .obj/.mtl files.
For me the most interesting feature was the use of material transparency and color blending which will allow me to work on my thesis on advanced visualization techniques. Nevertheless, the Image recognition suite can allow some increased accuracy if the API is flexible enough to modify the positioning of the GPS and even replacing the positioning when combining some 3d existing models.
...
