The Merlin Bird Photo ID program is designed to recognize birds quickly and easily. This new app, developed by researchers at Cornell University, allows backyard naturalists and avian aficionados to identify birds in photographs.
Pictures are selected within the application, and the user encases the bird to be identified inside a box. The user next selects an approximate size of the observed bird, its base color and what the animal was doing when it was spotted. Positioning of the bird is deduced from the user pinpointing the bird's eye, bill and tail.
"Merlin asks you the same questions that an expert birder would ask to help solve a mystery bird sighting. Notice that date and location are Merlin's first and most important questions. It takes years of experience in the field to know what species are expected at a given location and date," the Merlin Photo Bird ID team wrote on their website.
The app then compares the appearance to its database of around 400 birds found around North America. When a list of candidate birds is generated by Merlin, the various species are presented to the user, along with high-quality photographs and recordings of the sounds the animal makes.
Bird data was collected from 70 million observations that comprised the eBird project, which recruited amateur ornithologists. Thousands of backyard bird lovers contributed three million descriptors to the app to improve its performance.
The application offers the shortest possible list of candidate species from which users can select the bird they spotted, based on the information provided and the appearance of the targeted bird in the photograph.
Merlin was developed by the Visipedia research project, in cooperation with investigators from the Lab of Ornithology at Cornell University.
"It gets the bird right in the top three results about 90 percent of the time, and it's designed to keep improving the more people use it. That's truly amazing, considering that the computer vision community started working on the challenge of bird identification only a few years ago," said Jessie Barry, Merlin Project Leader at the Cornell Lab of Ornithology.
Future versions of the Merlin application could ask users for the type of bird in the photograph, whether it's a raptor or a songbird. Until then, the machine-learning capabilities of Merlin will allow the application to continually become more accurate over time.
The U.S. Fish and Wildlife Service (FWS) estimates there are around 41 million backyard birders in the United States.
Merlin is available for free on the Apple App Store and in Google Play, although the download for the app is over 600 megabytes for Android.
