Log in or register to post comments

Cylinder vs Image target

April 3, 2019 - 7:39am #1



I'm trying to track a bottle label and can't get satisfying results.


When I use image tracking I'm getting really good image detection. Like 90% of the time it catches the label straight after pointing the camera on it. But it has horrible tracking afterwards. My virtual object is shaking and dropping so much that it's unacceptable, even with smooth camera script.

When I use the same label but as a cylinder target I'm getting ultra bad detection, close to 5% succes rate, which is unacceptable as well. However when it eventually does detect the label, the virtual content sticks to it like it was nailed. I can rotate the bottle, shake the camera etc and it will still be right there.


I've noticed that when I upload the image to my cylinder target it always gets scaled down and pixelated, like it was 50% of the native resolution, it looks like this both in target manager and in unity scene. Is it only scaled so it doesn't take up much space on vuforia's hosting, or is it the image that is being tracked afterwards - which looks a lot different to the real label. I've tried doubling the resolution of my image before uploading and it doesn't change a thing, always getting scaled to ultra bad.


Does anyone have hints?



Cylinder vs Image target

April 5, 2019 - 9:04am #2


Strange that you're seeing such a noticeable performance difference between the two as they tech behind them works on similar principles.

Can you share the target (source image) with me for further analysis?


Vuforia Engine Support

Log in or register to post comments