I fired up an instance of DOODS in Docker and oh boy did it give me a whole lot of false positives. Way more 'people' in an image than there were. I'll give it some more images later today or tomorrow and see if I can tweak it at all.
If you're still struggling with this, I found that I received the same error when I had VISION-DETECTION set to TRUE instead of True. For some reason having it in all caps made it not work for me.
I'd say that's a bit too deep as you'd need SMTP authentication details, etc. If I wanted that info likely look to MQTT and Node Red.
You can do this within BI. I have it to setup to send to PushOver, but they can also be sent directly to an email account.
So you have a 24/7 recording of everything, not only the alerts.
You can have AI TOOL send you the image to your Telegram app. Don't think BI can do it.
Very much so, but making the move to becoming open source. VorlonCD's fork of the original AI Tool is the one most of us use.
I may be missing the obvious but I don’t see the benefit of doing this. If my wife started seeing this in her notifications she wouldn’t be to happy.
Agreed, if you're going to make adjustments you are in the application anyway. And you can review multiple confidence levels all at the same time.
When I review images in AITool and notice my Cat is being detected as a dog I noticed I can right click and have the "Prediction details" option, is this where you adjust confidence levels? I'm sorta confused as to where to tell it this isn't a dog but its a cat or sometimes it detects my dog thats 35 lbs as a person on cameras that I have it to detect both dogs and people, on another camera where I have it detect dogs and cats...sometimes the Cat if detect as a dog and I need to specifically detect the dog when he eats cat food.
