Hello! I am currently working in one of the projects involving hazard detection in construction site using OAK-D Lite. I wonder if is it possible to implement pose estimation (for fall detection), crowd counting, and equipment safety (forking Gen2 Human-Machine safety) all at once? And also how can I able to stream it to the web server?

    Erwu117 And is it possible to measure the two classes by using depth sensor and yolo as nn?

    • erik replied to this.

        Hi Erwu117 ,
        It would be possible, but it would be likely very slow, as you would want to run pose estimation + object detection + crowd counting all at once. Just runs at about 1FPS, adding the two other NNs would then decrease FPS further.

        So I would suggest keeping an eye on the OAK Series 3, which should achieve about 5x AI performance, which would allow you to run all these features at decent FPS.

        For streaming to webserver, see options here.
        Thoughts?

        Thanks, Erik

            erik Noted but where can I find the source code for fall detection and crowd counting under one main program?

            • erik replied to this.