[8fingerlouie]

I use a homegrown solution consisting of Raspberry Pi Zero W's with cameras, using the motion (https://motion-project.github.io/) daemon, saving their captured video to a NFSv4 share on a server.

The shares on the server are grouped so that each individual camera has a subdirectory of a parent directory, which is in turn shared by Syncthing to another local mirror and a remote mirror.

A python script runs on the server, using Pyinotify to detect new files, and using TensorFlow to do basic object detection, and adds bounding boxes to videos where it detects humans.

Finally a notification is sent through Pushover via MQTT (Mosquitto) when a person is detected, along with an image and a camera name and timestamp. It does presence detection by pinging our phones, so notifications are only sent if nobody is home.

If i should do it all again i would probably just buy a couple of Unifi cameras and a Cloudkey Gen 2 Plus and be done with it :)