Detecting falls using DNN, looking for advice

edit

Hi All, Not sure if this is the right place to ask such question, but if not, I am sure you will point me in the right direction. I plan on building a reasonably simple application that can detect falls. I thought about using openpose but it seems to be too slow and I am aiming this to run on relatively cheap hardware (google aiy?). There are faster solutions using mobilenet ssd like https://github.com/ildoonet/tf-pose-e... which might be ok but... here comes the questions: 1. would training mobilenet ssd to detect standing, lying down or on four (half fallen) person would be difficult? 2. would this be a faster approach then detecting pose? 3. any ideas for datasets to use?

regards, Moshe