Tried Frigate. Not switching, but makes me want to improve my BI setup

[The_Penguin]

So after reading some comments here and there about how good Frigate is, I thought I'd try it. Can't say I like it much, but holy crapola! Out of the box, the object and facial recognition were awesome. On a Gen 7 i5 box.
Really hate the UI and the configuration via YAML. Wow, I'd hate to be a non-techie trying to set that up.
So I want to focus on my current BI setup and get it working more like how I want it.
My hardware really should be good enough. i7-10700K 32G RAM. A 6GB RTX3050.
What I don't like is the object detection missing things. I have motion detection set very sensitive because of this, so I get lots of blowing spider webs and tree shadows. If it didn't miss detections, I could fix that.
Aside from hoping that v6 brings some improvements, I'm wondering what I can do with my AI settings to make it better. Maybe I'm missing something.
Currently BI is 5.9.9.73 Code Project 2.9.5 using YOLOv5 6.2 1.10.0
So what's the recommended setup for good AI for BI these days?

 

[wittaj]

Recommended setup for many is cameras with AI built-in.

If not then it is a lot of trial and error and not trying to do too much with one field of view.

The larger the object is in the field of view, the better CodeProject will work.

 

[The_Penguin]

Update:
I have played with Frigate for a while. Detection is very good, maybe not as great as I originally thought.
I'm really hoping there's some great improvements with BI 6, because - maybe I'm crazy, but I really don't like Frigate. I see people say they left BI for Frigate and love the UI, and I don't get it.

 

[wittaj]

If you don't have cameras with AI and you want our help on dialing in your CodeProject, some screenshots of your settings and field of view is helpful.

In some instances, it is simply trying to do too much with one camera. In other instances it is a simple setting adjustment like adding more contrast.