Repository navigation
Frigate+ for baboon detection! Monkey candidate? #24971
Mwanafunzi
started this conversation in
General
Replies: 1 comment 5 replies
|
Wow - I get excited when I see a fox or a badger! Baboons are quite something but I appreciate they are a hazard in many ways though. I have no experience of training for things which are not yet labels (for example badger has not yet been promoted from candidate to label) but I have found Frigate+ to be very effective. I now get nearly 100% accuracy for foxes for example. Good luck! |
5 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment


Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
I remotely manage a Frigate 0.17 setup in E Africa, where baboons raiding crops are their biggest problem. The default model scores baboons as person or bird at > 0.65, so they’re mostly missed and confusing with human would be a PITA for alerting. To my eye, I'm surprised at the misclassification. So I'm considering Frigate+ and assuming I can train-up monkey detection?
Before I start - is monkey mapped to an existing label during training? Would annotating baboons as monkey teach the model to treat them as background until the label goes live? If so, should I hold off on annotating them, or label them some other way?
Also, I’d like to register interest in monkey being promoted to a supported label. Primates are a real crop and security problem in much of Africa and Asia, and I can contribute a steady supply of annotated images.
All reactions