App Helps Amateur Birders Identify Like the Experts
by Liam Farrell
For the curious hiker and birdbath watcher who spot an unusual feathered friend and want to know more about it, the slogan is true: There’s an app for that.
Birdsnap is the latest species identification app from UMD computer science Professor David Jacobs, who along with researchers at Columbia University developed the earlier Leafsnap (also in partnership with colleagues at the Smithsonian Institution National Museum of Natural History) and Dogsnap apps.
“We’ve learned how difficult it is, even for people with training, to do species identification,” Jacobs says. “We’re so used to search engines, and this is a search engine model.”
Birdsnap, which has been downloaded 25,000 times and is available at birdsnap.com, builds on the work that Jacobs has done in the field of identifying human faces. With lots of examples, technology can essentially learn the geometric standards of a species and narrow down choices for a user.
“We can train an automatic system to extract a representation of what a beak looks like,” Jacobs says.
When a user uploads a photo of a bird, the database searches through 500 of the most common North American species and makes suggestions based on its appearance and the likelihood of seeing it in a particular area. Links are also available for bird sounds. Jacobs says this is all helpful for novices who may have difficulty with the more standard field guides.
At this point, creating each of the identification apps necessitates rebuilding the wheel each time. Jacobs eventually wants to create a single tool that can use more universal techniques and shift between identifying things like shells, insects and weeds.
“There’s a tremendous desire by biologists for these types of tools,” he says.
