Body Detection Method

[omroali]

Picking up and estimate the distance of the victim.

I've been looking into how to pick up bodies. I am assuming you've come across yolo
https://www.pyimagesearch.com/2018/11/12/yolo-object-detection-with-opencv/ and so far seems like a nice solution to me.

I've been testing it out on my laptop. It reliably analyses a single image in about 0.5 seconds and doesn't seem to be very intensive on my CPU (usually 1-2%) I'm not sure how this will relate to the Pi4 though.

Assuming this is a reasonable solution, I'm still unsure how to get it to recognise distance. I might be able to get it to pick up the size of the surrounding box and relate it's height with distance. Alternatively, maybe the real sense might be able to give some indication of the distance to the victim, though I don’t really know much about it.

If this is not a reasonable solution then, I'm not really sure of any other good methods that wouldn't be too intensive for the moment. Would be happy to try stuff out if you have suggestions.

(also if this is the wrong way to use the issues section, feel free to correct me)

[fsherratt]

On this repo there's not really a wrong way to use issues.

It will only run slower on the pi and 0.5 seconds probably a little slow. Any ideas which part is taking the time/obvious optimisations (lower resolution)?

It would definitely be possible to use the real-sense to add more information but with the timescale we've got it's probably only going to be possible to use a RGB image. I think primarily we just need this to produce a picture of the injured person to send to the MK:SmartHub

Full rules are here if you want to see exactly what we need to do https://github.com/fsherratt/ERL_SmartCities_2019/wiki