I am testing your SDK for a potential project we are working on.
I want to use the SDK to recognize a set of products placed on shelves, and do the calculations of total quantity of each product.
I have tested the demo app, I used the Multiple Targets Image tracking, and I was able to detect the products.
But the problem is when using the camera, a simple moves will result in the same target can be tracked more than once, which results a wrong total quantity.
Is there any way to use an image instead of the camera capture feature?
Later on, we need to compare two different images of the same shelf taken in different days, this will be used to check the difference in quantity.
I am currently using the native android SDK, with Android Studio.
Hope you can assist in my query,
I don't understand why you would need an image as an input and I don't see why moving the camera and recognizing another instance of a target would result in an incorrect count.
If all you're interested in is the quantity of targets, be it the quantity of distinct targets or the quantity of a specific target, all you need to do is keep track of when a target is recognized (+1) and when a target is lost (-1).
If you take the name of the target into account, you will end up with the number of actively tracked instances of each target for each point in time. I don't see what information you could possibly be lacking at this point.
Thanks for the quick reply.
The app will be used to compare Planograms, we need to capture different Planograms pictures and get the quantity difference of all the products.
If you would like to forgo the camera stream and input these planograms directly you will need to use our plugins API; specifically, the input plugins API. It allows supplying a different frame input source for the SDK (still image or video). The relevant documentation can be found here (iOS) and here (Android).
We, of course, have sample code in our sample app for this set features.
Is this what you needed? Is there something left unclear about the counting process?
I will look into the plugins API, but does it support multi target?
I have no issues with the counting as long the camera does not move, Maybe by using still images things will be better.