Hacker News new | ask | show | jobs
by barkingcat 2304 days ago
Because it doesn't require multiple Kubernetes clusters spun up on 5 different geographical regions for fault tolerance with 3 different vendors to run.

It works too well so people would rather downvote the advice for which kind of salt to use.
1 comments
smoyer 2304 days ago
That's actually a great idea ... I'm going to create clusters of Bug-a-Salts on PTZ platforms, write a custom Kubernetes operator and then use Kubernetes to orchestrate their target acquisition and elimination.
link
barkingcat 2304 days ago
you might run into latency issues while using existing cloud platforms to run the AI stack for image recognition via the 50 webcams you need to set up to provide coverage around the house.

It's better to build a hybrid cloud whereby the image recognition code runs on tensorflow locally on 10 x RTX 2080 super's (for low latency inference) and then go on to issue commands via the kubernetes cluster for elimination.

link