Start a new topic

Image Recognition Problem

I have had success with using wikitude to recognize several different targets.  However, I have a particular target that I am having a lot of trouble getting recognized.  I find it strange that I use several images that get two stars, but cannot get those exact same images recognized.  I have tried using a lot of images.  I have tried just one image.  I have attached a wtc with one two star image that I am having trouble with.  Any help would be greatly appreciated.

wtc
(33.1 KB)

Hello Heather,

I just tested your .wtc file by replacing it within our sample 'Image On Target' and every time I was able to successfully scan and recognize the target. Do you think that maybe you can send over a video demonstrating your issue so that I can investigate further? Also, you can refer to this documentation section for general guidelines on 'Best Practice for Image Targets'.

Thanks
Eva

 

We actually are trying to use these images to target a small plush key chain.  We have actually gotten some of the images in this new wtc to be recognized.  However, we have not been successful in getting the original object to be recognized.  I am attaching the new wtc.  Do you think it could be possible to get the original object to be recognized with some amount of images?  Any advice would be greatly appreciated.  We have been successful using images to have other small similar objects recognized.  We have just really had trouble with this particular object.

wtc
Hello Heather,

If I am not mistaken you want to do Object Recognition. For this, you do not need to scan 2D images but the object itself since our new SDK 7 enables now Object Recognition. You can find more information here:

Thanks

Eva

 

We were able to get the object to recognize withimages using your cloud based system.  Do you know the difference with the cloud based system versus your local system for the recognition functionality?  We were hopeful we could address the problems and move forward with the cloud based system.  However, we just read that the cloud system limits the amount of scans to a 1,000,000.  This is an unacceptable limitation.  Any ideas on how to get a better scan locally similar to result in the cloud system?

Hello Heather,

I believe you already had a Skype call with my colleagues Nicola and Paula and you are also in contact with Paula via email. I will paste her answer here for other forum users as well:

"In Cloud recognition, the computation is done on a server with a more powerful hardware than on mobile devices. Therefore we can afford to perform more complex recognition algorithm which can yield better results comparatively. Additionally, with the launch of SDK 7 there were significant improvements in the cloud recognition service, making the recognition and tracking in the cloud serivce even more robust compared to previous versions."

Thanks
Eva

 

Login or Signup to post a comment