Hi jsiic
Yes, the Oak-D devices run the pipelines on their own hardware. This is not limited to standalone mode. When you use the node.MobileNetDetectionNetwork in the API, the processing is done on the Oak-D device itself. This is because the Oak-D device has an onboard Vision Processing Unit (VPU) that allows it to process AI and computer vision tasks directly on the device.
On the other hand, if you were to send the Oak-D video feed into a MobileNet script written on OpenCV, the processing would be done on the computer. This is because OpenCV runs on the host machine (your computer), not on the Oak-D device.
The advantage of running the processing on the Oak-D device is that it offloads the host processor from this work. For example, in the common use case of object detection from a 12MP image, your host is now dealing with a 24 Kbps stream of what the objects are and where they are in the image, instead of a 2.1 Gbps stream of video. This results in a significant reduction in data your host has to deal with.
Thanks,
Jaka
