Log in or register to post comments

Problem for recognition of two images similar

August 1, 2012 - 3:53am #1

Hi all,

I have a problem with the recognition of two images most similar to each other. Vuforia can't determine precisely which image is between the two, and constantly changes the  3D content associated with the image target.

It's possible set a minimum threshold for recognition, such as a minimum percentage of corrsipondeza between live image and dataset, a target image?

Thanks,

Alessandro

Problem for recognition of two images similar

October 9, 2012 - 8:58am #12

Hi DavidBeard,

unfortunately I did not have access to the beta version of the cloud. Can you tell me if the cloud provides a service that allows to know if I have loaded an images target similar to a previously loaded image?

Thanks,

Ale

Problem for recognition of two images similar

October 9, 2012 - 8:58am #11

Hi DavidBeard,

unfortunately I did not have access to the beta version of the cloud. Can you tell me if the cloud provides a service that allows to know if I have loaded an images target similar to a previously loaded image?

Thanks,

Ale

Problem for recognition of two images similar

August 7, 2012 - 5:52pm #10

We're introducing a web service ( Vuforia Web Service ) that may able to address your requirements - https://ar.qualcomm.at/content/vuforia-web-services.

 

Problem for recognition of two images similar

August 7, 2012 - 12:47am #9

I do not know beforehand whether there are two similar images. I want an alert that tells me that the immgine framed could be very similar to another image of the dataset.

And maybe you also know the detail of the image similar.

Problem for recognition of two images similar

August 6, 2012 - 10:55am #8

If you know that there are similar images in the dataset, can you simply present that message when any of them are detected? - i.e. identify them by their Trackable name.

Problem for recognition of two images similar

August 6, 2012 - 10:49am #7

In my use case I could have two similar images, the SDK can not stabilize the content to display for that image. I wish I could have a warning saying that the live image is very similar to another image of the dataset.

I hope I was clear :)

Alessandro

Problem for recognition of two images similar

August 6, 2012 - 10:49am #6

In my use case I could have two similar images, the SDK can not stabilize the content to display for that image. I wish I could have a warning saying that the live image is very similar to another image of the dataset.

I hope I was clear :)

Alessandro

Problem for recognition of two images similar

August 6, 2012 - 7:30am #5

Not from the API, there's no measure of image similarity provided. Can you describe your use case? There may be an alternative approach.

Problem for recognition of two images similar

August 6, 2012 - 3:34am #4

Ok, as I thought... "There is no way to set a precision threshold through the API."

But, there is a way to warn the user that the image target is similar to another image of the dataset?

Problem for recognition of two images similar

August 1, 2012 - 7:39pm #3

Similar images are challenging because the Tracker needs to be able to recognize a physical target image even if it's not a perfect representation of the image used to generate the Trackable data - this is how the Tracker can continue tracking even if the target is partially occluded, for instance. And even changes in color between very similar images may not be sufficient to distinguish them because the Tracker utilizes a grayscale version of these images - so if they look the same in grayscale, they'll likely be confused. You can observe the grayscale versions in the TMS interface after generating a trackable, or by clicking on the image in your project view.

My advice is to either alter the images so that they don't match in grayscale, or add a FrameMarker border to each and track those instead. There is no way to set a precision threshold through the API.

Problem for recognition of two images similar

August 1, 2012 - 4:30am #2

Do the ImageTargets look similar to each other? I once had the same problem and changing one of them significantly solved the problem.

Log in or register to post comments