I'm interested in multiple devices per host. What I want is to use multiple cameras as a single camera to perform functions like an object tracker. (For example, several cameras capture a continuous space in sequence, making it seem like a single camera is detecting objects.) I've looked at Luxonis docs and the gen2-multiple-devices on GitHub, but it seems like each camera is used independently and not as one. Did I misunderstand? Unfortunately, I only have one camera at the moment so I cannot test the code myself. On the other hand, is it possible to use multiple cameras independently? For example, can two cameras in one host, each looking at a different location, perform Cumulative Object Counting independently ?

Hi june You can't really make it look like it's a single camera, since different cameras have different positions and perspectives that you cannot perfectly align unless the objects are very very far away. You can still combine the two feeds together side-by-side in cv2. The bounding boxes should work as well, you just need to offset them by first camera width. This would enable you to display results in a single window, but there will be a distinct border between the two feed. Also, the bounding boxes will not work cross cameras unless you use a model that can process double the width. Yes. Thanks, Jaka

Using Multiple Devices as One or Independently

june

I'm interested in multiple devices per host.

What I want is to use multiple cameras as a single camera to perform functions like an object tracker. (For example, several cameras capture a continuous space in sequence, making it seem like a single camera is detecting objects.) I've looked at Luxonis docs and the gen2-multiple-devices on GitHub, but it seems like each camera is used independently and not as one. Did I misunderstand? Unfortunately, I only have one camera at the moment so I cannot test the code myself.
On the other hand, is it possible to use multiple cameras independently? For example, can two cameras in one host, each looking at a different location, perform Cumulative Object Counting independently ?

jakaskerl

Hi june

You can't really make it look like it's a single camera, since different cameras have different positions and perspectives that you cannot perfectly align unless the objects are very very far away. You can still combine the two feeds together side-by-side in cv2. The bounding boxes should work as well, you just need to offset them by first camera width. This would enable you to display results in a single window, but there will be a distinct border between the two feed. Also, the bounding boxes will not work cross cameras unless you use a model that can process double the width.
Yes.

Thanks,
Jaka