#TODO once it actually works!
As of December 1st, 2017, this is still a work in progress. The current plan is to use a security camera mounted on the corner of the house that can see the entire street out in front. The recorded video for the evening will be processed using YOLO Darknet (probably actually Tiny YOLO). We will write some code around YOLO to track the detected car objects and categorize them by how long they stay watching the show. We will publish this data somehow on our home network so we can have a realtime look at how many people are watching our show (perhaps even having our Echo dot be able to tell us how many people are in front of our house at any given moment).
The biggest concern right now is that the only hardware we have to run the video processing is either a Raspberry Pi or a five year old Zotac ZBOX nettop computer. Neither of these has a modern GPU or a particularly fast processor so they may not be able to keep up to process video in real time. Other Christmas light tasks have gotten in the way thus far but hopefully once the show is solid, we can get this cool feature figured out.