- My Forums
- Tiger Rant
- LSU Recruiting
- SEC Rant
- Saints Talk
- Pelicans Talk
- More Sports Board
- Fantasy Sports
- Golf Board
- Soccer Board
- O-T Lounge
- Tech Board
- Home/Garden Board
- Outdoor Board
- Health/Fitness Board
- Movie/TV Board
- Book Board
- Music Board
- Political Talk
- Money Talk
- Fark Board
- Gaming Board
- Travel Board
- Food/Drink Board
- Ticket Exchange
- TD Help Board
Customize My Forums- View All Forums
- Show Left Links
- Topic Sort Options
- Trending Topics
- Recent Topics
- Active Topics
Started By
Message
Ai/camera training
Posted on 5/27/26 at 5:48 pm
Posted on 5/27/26 at 5:48 pm
Anyone doing this? I am thinking of getting some cameras and training ai to recognize certain species of birds with a high degree of accuracy, as part of a larger project.
3
Posted on 5/27/26 at 8:18 pm to DonJuanDaMiles
Check out rigate
Posted on 5/27/26 at 10:11 pm to Vrai
Frigate?
I did another thing with Verkada for cameras and remote gateway and an ai company for the other stuff
I did another thing with Verkada for cameras and remote gateway and an ai company for the other stuff
This post was edited on 5/27/26 at 10:12 pm
Posted on 5/28/26 at 7:02 am to DonJuanDaMiles
Posted on 5/28/26 at 7:21 am to j1897
Maybe OP doesn't want to spend $200, or wants to challenge himself to put it together on his own.
Posted on 5/28/26 at 7:29 am to TigerFanatic99
quote:
recognize certain species of birds with a high degree of accuracy
Training this from scratch is going to take hours and hours of data preparation, and then it's gonna make a 5090 cry while it trains. 200 dollars is insignificant.
I would use an already trained model TBH
Posted on 5/28/26 at 7:50 am to DonJuanDaMiles
I use ultralytics with mostly kaggle and Roboflow datasets.
Curating datasets is the hard part. You’re talking 10k images, all augmented for your image pipeline.
Choosing a model is just a matter of matching resources as long as you’re using a modern model. Everything good except the heavy RF-DETR is copyleft so not commercial viable without license but you can use them at home.
Curating datasets is the hard part. You’re talking 10k images, all augmented for your image pipeline.
Choosing a model is just a matter of matching resources as long as you’re using a modern model. Everything good except the heavy RF-DETR is copyleft so not commercial viable without license but you can use them at home.
Posted on 5/28/26 at 8:38 am to Vrai
Frigate is absolutely top tier for their corse to fine motion augmented detection pipeline and object tracking. Their automation API is best in class by a long shot.
For OP’s use case, they fall behind due to their strict stance on permissive licensing. There just aren’t that many good vision tools out there that are permissive until you get quite heavy like RF-DETR or Qwen VLMs that are not suitable for surveillance and tracking on consumer hardware. Their bird classifier is a tiny dinosaur on par with what you’ll get with a commercial feeder product.
Fortunately, Frigate is easy to modify and already unofficially supports modern models like Ultralytics.
For OP’s use case, they fall behind due to their strict stance on permissive licensing. There just aren’t that many good vision tools out there that are permissive until you get quite heavy like RF-DETR or Qwen VLMs that are not suitable for surveillance and tracking on consumer hardware. Their bird classifier is a tiny dinosaur on par with what you’ll get with a commercial feeder product.
Fortunately, Frigate is easy to modify and already unofficially supports modern models like Ultralytics.
Posted on 5/28/26 at 10:33 am to Dallaswho
Roboflow was the ai company I was talking to.
Posted on 5/28/26 at 10:37 am to TigerFanatic99
quote:
Maybe OP doesn't want to spend $200, or wants to challenge himself to put it together on his own.
This isn’t for a bird feeder and I talked to a CEO of one of those companies and have been talking to Cornell who has the Merlin bird ID app. The best trained models for bird recognition still aren’t performing that great in the application I’m looking into for recognizing known disease carrying species.
So, the training would be to fill in those gaps or to figure out a way to make them more efficient.
Page 1 of 1
Popular
Back to top
Follow TigerDroppings for LSU Football News