A standalone PIR is a tenner. Build something temporary (but not actually dangerous...) and test it. I feel anything with computer vision or wireless is over complicated and will suck in the long run.
I have a Hue PIR that I'll test it out with first. There's nothing smart about the PIR part so should hopefully give me a decent idea of whether any of the positions would be viable.
A standalone PIR is a tenner. Build something temporary (but not actually dangerous...) and test it. I feel anything with computer vision or wireless is over complicated and will suck in the long run.