Being able to correctly identify which bird comes to your feeder is one of the most rewarding experiences of birdwatching. To ensure Bird Buddy successfully recognises species in all types of habitats, lighting, and wildly varying weather, we’ve been feeding those images from the “test drive” camera network to our AI (Artificial Intelligence) technology so that it can clarify just who is chowing down at that feeder. So far, we have received over 500 thousand images of around 50 to 60 different species, which is a remarkable achievement, and we are so thankful for these amazing contributions. Using the photos from our test network, we have been able to significantly improve existing models for image identifications, and eventually, with the amount of data we have coming in, we’ll be able to create a model that the best ornithologists will struggle to compete with.

Being able to correctly identify which bird comes to your feeder is one of the most rewarding experiences of birdwatching.

To give you an idea on how we train our AI classification models—all images go through two stages of processing to ensure the best result. On the first step, we determine whether the image is “interesting” or “not interesting” (a bird is in frame and in focus, at least one eye and the beak is visible). We have now gone through all 500,000 initial images and classified them according to that criteria and learned that the images that are normally interesting to a human, are also ones that AI will have a much easier time doing inference of the species on. Yay! About a third of all the images come out as interesting and our amazing in-house ornithologists are going through each, to classify species captured on the photo. All of this is fed into a model with a small group of classified images left for validation. The more varied images we get in the better the model.

So, we have those initial most common species covered with our improved model, and we want to expand the coverage to as many species as possible. In order to do that, we’re going to need more images, and this is where YOU can come in.

In order to create a truly top-notch AI classification model, we’ll need approximately 5,000 images of each species taken from every angle and in all conditions. So, buddies, we are now asking all of you bird lovers out there to partner up with us and send in any photographs or footage you have—amateur and professional alike. We’d love to see all those images, the blurry-winged ones, the half-a-head-in-shot ones, the tail-feather-only ones, all of them. The more varied images we get, the better the education of the AI.

For our AI to learn more, we’re going to need more images, and this is where YOU can come in.

From this valuable input, not only do you get to rejoice in knowing for sure just who that little guy is eating your seeds, we can help more people learn about, care about, and appreciate birds, and at the same time find ways to help them.

‍To find out more and upload your images, click this link, and thanks again for being there for the birds.