Most of the time, people set up bird feeders because they are looking to see some birds. You can’t be staring out your window all the time, though, so you are bound to miss at least a few of your feathered friends who drop by your birdhouse. Smart bird feeders allow you to see every member of the avian community who comes for a visit. Some of these designs can also identify the birds they take photos of.
There might be a few bird watchers who recognize every fowl they see at a glance, but most of us need a little help with the more exotic varieties. Luckily, the software Stockman programmed for his smart bird watcher uses the free tier of the Microsoft Azure Custom Vision machine learning service. This pulls from similar images to help bird watchers identify their winged guests.
Stockman explained that he trained his model by pulling 20-25 photos from Google Images for each popular bird species that would come to his feeder, using this information to train Azure to tag each photo by species. Stockman admitted that this method isn’t 100% foolproof, but it does help with the identification and categorization of local species.
Of course, anyone setting up one of these smart bird feeders will have to train their own modules to recognize the birds that live and migrate through their area.
