It is necessary to use the Target Management System to turn your images into trackable objects. The system does not make any improvements to your image, it is up to you to provide an image with sufficient detail and contrast for tracking. See the Developer Guide > Target Management System > Create Image Targets for more details.

The system will process your image and find a set of features that can be used for tracking. It provides a rating (from 1 to 5 stars) on the quality of the image for tracking. Click on the trackable to see a detail view with the feature points overlaid on top of the image. This can help you figure out which parts of your image are feature-rich and will track well, and which parts are lacking well-defined features.

- Kim

It would be great if there was an offline version of the software that provides the tracking data from source images.

From experience so far, it seems that the rating that the TMS gives isn't entirely indicative of the TRUE trackability of the target image. Some of my 4-star images completely suck and are barely trackable, wheras some of my other 4-star images are perfect! We need a solution that allows developers to very quickly try new images in real-life application. This business of uploading an image, having it analyzed by qualcomm's servers, and then downloading it back again is nothing more to developers than a pain in the ass and a bit of an obfuscation of the image processing mechanics that go into the process.

I just was thinking about it, and join this advice. Qualcomm Developers, please, make an API to upload image and get .dat file!!! I can even make an opportunity to rate trackables quality in my app. So you'll get tons of results.

The main idea is to make a photo of some object, and instantly see it's AR application. So, i can make a funny app that take photo of your friend and add some funny effects (big lips or ears, add pirot hat or smth). You can take photos of your coworkers in office, make some funny impruvments, and go seeing different people. BR

P.S. it could be a good idea for startup to make an online API service for making .dat trackables on photo-images, it could be imagemarking.com, it's free now))) Who will make? I have no resources to make it alone.

I agree with hybridbuzz and stanislap, an API or an offline version would be great.

arcv wrote:

I agree with hybridbuzz and stanislap, an API or an offline version would be great.

I even believe that it will be a crucial factor for some developers. Imagine you wanted to create trackables of objects in a car factory. You'd never be allowed to send them to external parties. Hence, trackable creation would have to be feasible without any outbound IT communication. Even a web service provided by an external party would not be sufficient from a security point of view.

Nevertheless, I also believe that there is a strong demand for a web service for trackables creation. Not all companies have such strict regulations.

I would like to cast my vote too. Qualcomm, please make the trackable generation system offline! Why is it only online? It is a super secret algorithm you are using? Come on!

the most interesting thread on this forum i think.
any answer from qualcomm here would be a great help in deciding which ar-framework to use.

any update on this thread?

I've decided to bump all of these threads until we get an answer as to why this can/will or can't/won't be done.

I can't send my clients' proprietary, unreleased, and confidential brand imaging to another company for any purpose.  We all love the SDK and it works pretty darn well after .dat file creation.  The fact that you have to manually upload data to a black box is a roadblock for any business application of this technology.

PLEASE, at least address the issue of you're not going to satisfy our request.

Hi,

As many of you already said, it would be nice to have an offline or integrated function to get the trackable data. However I also want to cast a vote in favour of Qualcomm current method. As you already known, Vuforia is new project in current grow and development, I think that Vuforia currently needs an online system in a way that every image we submit can be used by them to review, analyze and improve the SDK, so by submitting them online we are helping them.

Maybe in a future once they get enough information they could plan to release an offline method.

Besides, bear in mind that computer vision and imaging recognition processes are computer-intensive tasks, based on that, the method they use to extract the images features could take hours or minutes to process in our computers, at that point having an online server release us of that tasks and let them do the work for us in a more powerful and faster fashion.

Don´t take my wrong, an offline system would be great, let´s give Vuforia a bit more time to work on this amazing SDK.