Buzz
While pet-sitting for a friend last October, I stumbled upon an unusual bird perched on a car hood in Brooklyn during a walk with the dog. After noticing it hadn’t moved for an hour, I decided to relocate it to a nearby park. However, I faced a dilemma: how could I rescue a bird without knowing its species?
If only I had known about Birdsnap. Developed by researchers at Columbia University and the University of Maryland, this digital guide can recognize 500 North American bird species using just a smartphone photo. The steps are straightforward: snap or upload a photo, mark the bird’s eye and tail, and let the app generate possible matches.
After releasing the bird in the park, my birdwatching friend informed me that another enthusiast believed it was a young Virginia Rail, a reclusive marsh-dwelling bird. Birdsnap confirmed this identification instantly. (Its appearance on a Brooklyn car hood remains an unsolved mystery.)
The concept for the app originated from Columbia Computer Science Professor Peter Belhumeur and University of Maryland Computer Science Professor David Jacobs, who recognized that facial recognition technology could be adapted for species identification. Facial recognition algorithms analyze similarities between facial features, such as comparing noses to noses and eyes to eyes, as explained by Columbia's Engineering Department. In Birdsnap, 17 specific parts of each bird species are labeled, enabling the app to analyze these features, match them against its database, and identify visually similar species from uploaded photos.
"Birdsnap is particularly groundbreaking because it not only excels at species identification but also pinpoints the specific bird features the algorithm uses for recognition," explained Thomas Berg, a Columbia Engineering computer science PhD candidate. "The app highlights these key features—known as 'field marks' among birders—on the bird's image, helping users understand what to observe in the future."
Beyond identifying birds, the iPhone-compatible app offers detailed descriptions of species, their vocalizations, evolutionary family trees, and visually similar relatives. It also includes range maps and data on sightings to enhance the birdwatching experience.
