Pose estimation is useful in a slew of real-time situations including worker safety (presence of hands next to other equipment) and even things like sports (for real-time correction, feedback), etc.
So supporting something like OpenPose (like here) would be useful in many use-cases.
Move to the how:
Implement the tensor parsing on the host side for MobileNetV2-PoseEstimation. Since this is MobileNetV2 based, it likely already runs on DepthAI, we just need to do the metaout decoding on the host along w/ display.
Move to the what:
Support MobileNetV2-PoseEstimation on DepthAI/megaAI.
Start with the
why
:Pose estimation is useful in a slew of real-time situations including worker safety (presence of hands next to other equipment) and even things like sports (for real-time correction, feedback), etc.
So supporting something like OpenPose (like here) would be useful in many use-cases.
Move to the
how
: Implement the tensor parsing on the host side for MobileNetV2-PoseEstimation. Since this is MobileNetV2 based, it likely already runs on DepthAI, we just need to do the metaout decoding on the host along w/ display.Move to the
what
: Support MobileNetV2-PoseEstimation on DepthAI/megaAI.