OAK-D for calculating distance between people

julienbayle

Hi there, I'm trying to know if there was an example somewhere that would help us to start with a Oak D devices that would detect people, and ideally send x,y,z data to the computer with which the cam would be linked to.

Just trying to figure out how to design such an architecture.
Distance Calculations could be done on the computer. I would only need a global mean.

Any ideas or words about this would help a lot.

Best regards,
Julien

erik

Hello julienbayle , we actually have this exact demo - social distancing demo. It detects people, sends spatial coordinates back to computer, and from there is calculates distances between people.
Thanks, Erik

julienbayle

erik checking it asap. Thanks a lot Erik. Especially, I'd like to "retrieve" and parse coordinates as text over udp or something like this. Ideally in openframeworks based exe, or max msp. I'll let you know.

julienbayle

erik hi Erik,
I'm stuck here.
Would you mind to help me to set this up for just sending coordinates as osc/udp message?
I can't figure out how to do this.
Actually, in the social app demo, I'd like to have only something sending messages, no calculations, or any gui, just firing this to local host on a specific port.

erik

Hello julienbayle ,
so I would add that logic here, instead of appending the dict to the bboxes. For UDP packet sending, you could check solution here. I hope that helps.
Thanks, Erik

julienbayle

erik thabks a lot testing this asap.
Which part. And which files would I alter?
Is it the main.py where everything goes considering the social distance example?

erik

julienbayle you could just alter the file that I linked to. You could remove most of other (unused) logic. The main file that you would need is depthai_utils.py (which constructs pipeline and receives spatial coordinates from the device), other files are for Covid distancing and you could remove them. So you could just alter depthai_utils.py a bit and only use that.
Thanks, Erik

julienbayle

erik got it.
Thanks a lot. In that case, it wouldn't create the ui window etc? Actually I wouldn't need that

And actually, I'd love to code an example using cam data DIRECTLY in other framework.
I'm thinking about max 8
Or touchdesigner.
I think there is a sdk somewhere 😀

julienbayle

Thanks I understand this.
It still doesn't work as I'd like.

I initiated the sock etc.
and put
MESSAGE = int(detection.xmax * width) sock.sendto(int(detection.xmax * width), (UDP_IP, UDP_PORT))
after the bbox.append() ...

but it doesn't seem to fire osc message.
I don't get it. code is ok
sock defined and assigned globally
after the import:

UDP_IP = "127.0.0.1"
UDP_PORT = 6666
MESSAGE = "!"

sock = socket.socket(socket.AF_INET, # Internet
                     socket.SOCK_DGRAM) # UDP

julienbayle

But the main problem is: I'd need to have distances as a message like:
distances 0.3 0.1 9.2 ....
if I have 2 persons detected : 1 distance (one float)
if I have 3 persons detected : 3 distances (three floats)
etc
actually I'd even need "just" a mean of distance.

Where should I do that in the code ??

erik

It looks like your UDP client isn't working as expected. Could you try to develop UDP packet sending without the depthai code? After that you can copy it back to depthai code. Regarding the floats, you can just take spatial coordinates - they are already there in the depthai_utils.py script.
Thanks, Erik