• Subscribe

Bird photo ID: Birders team with artificial intelligence to solve mystery

Speed read
  • Merlin Bird Photo ID lets you join in the science fun and identify birds from uploaded photos.
  • Taps into 70 million North American bird sightings and annotations available in the eBird database.
  • 400 bird species identified.

You know how it is: You’re on a late-morning hike and a bird pops out of the bush next to the trailside. You’ve never seen anything quite like it — but, unfortunately, there’s no ornithologist at hand. So how can you figure out what it is? If you’ve got your smartphone, then try out the Merlin Bird ID app.

Users helped train Merlin to recognize 400 bird species, including the Blackburnian Warbler shown here, by clicking on parts of the birds to provide information to the computer. Courtesy Christopher L. Wood.

“It’s kind of like a birding coach,” says Jesse Barry, Merlin project leader at the Cornell Lab of Ornithology. “Think of Merlin as a friend to give you cues for what you should be looking at and things you should be thinking about to make bird identification a little bit easier.”

Merlin, funded in part by the US National Science Foundation (NSF), comes in two versions. The smartphone app prompts you to enter characteristics and location information, and within seconds returns an identification, replete with pictures and audio samples. The web app takes it a step farther, letting you upload a picture to the Merlin website to aid in your quest to solve the avian mystery.

Merlin learned these identification skills by accessing about 70 million North American bird sightings and annotations available in the eBird database. With 250 million total observations, and 1.6 million more coming in each month, eBird is the largest data set on any natural living organism in the world.

The Merlin photo ID app, now available in beta version, is based on careful observation. Cornell ornithologists enlisted about 1,000 photographers to contribute 100,000 annotated images (photos where at least two different people have framed the bird and referenced 14 different parts of the bird).

The response from the community was immediate and strong, says Barry. “It was a really nice signal to us that this was a tool the birders wanted and that people were anxious to get help identifying birds.”

Hosted in the Amazon web services cloud, these photos constitute the underlying data set used to train Merlin to recognize 400 bird species. Merlin recognizes these images through fine-grained computer vision and convolutional network algorithms.

While Merlin can’t yet identify all photos taken from smartphones, as more annotated images are sent in by birders — from professional scientists to regular folks looking at birds in their backyards — the algorithms will improve and teach Merlin to recognize birds in photos of a lesser quality.

Cornell researchers have learned a lot already from the Merlin project, and have reported their findings to the research community in journal articles and conference presentations. At the other end of the data exchange, you get the answers you're seeking while walking the trail wondering about the pretty birds.

“We believe that helping people satisfy that natural curiosity and giving them an answer enables them to learn more about the species and to start thinking about caring enough to protect their habitat and build a stronger relationship with the natural world as a whole,” says Barry. “We’re at a point where humans need to redefine our relationship with the Earth and its resources so that we can reach a point of sustainability for protecting the environment and have it be a place where we want to live too.”

Now, if only Merlin could help you keep squirrels off your bird feeders.

A short introduction to Merlin Bird ID. Courtesy Cornell Lab of Ornithology.


Join the conversation

Do you have story ideas or something to contribute?
Let us know!

Copyright © 2015 Science Node ™  |  Privacy Notice  |  Sitemap

Disclaimer: While Science Node ™ does its best to provide complete and up-to-date information, it does not warrant that the information is error-free and disclaims all liability with respect to results from the use of the information.


We encourage you to republish this article online and in print, it’s free under our creative commons attribution license, but please follow some simple guidelines:
  1. You have to credit our authors.
  2. You have to credit ScienceNode.org — where possible include our logo with a link back to the original article.
  3. You can simply run the first few lines of the article and then add: “Read the full article on ScienceNode.org” containing a link back to the original article.
  4. The easiest way to get the article on your site is to embed the code below.