Communications of the ACM
Vision System Auto-Identifies Hundreds of Bird Species
Credit: iStockPhoto.com
In a breakthrough for computer vision and for bird watching, researchers and bird enthusiasts have enabled computers to achieve a task that stumps most humans — identifying hundreds of bird species pictured in photos.
The bird photo identifier, developed by the Visipedia research project in collaboration with the Cornell Lab of Ornithology, is available for free at: AllAboutBirds.org/photoID.
Researchers from Cornell Tech and the California Institute of Technology describe their results in "Building a Bird Recognition App and Large Scale Dataset With Citizen Scientists: The Fine Print in Fine-Grained Dataset Collection," to be presented at CVPR 2015, the Computer Vision and Pattern Recognition conference.
Called Merlin Bird Photo ID, the identifier is capable of recognizing 400 of the mostly commonly encountered birds in the United States and Canada.
"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," says Jessie Barry at the Cornell Lab of Ornithology. "That's truly amazing, considering that the computer vision community started working on the challenge of bird identification only a few years ago."
To see if Merlin can identify the bird in a photo, a user uploads an image and tells Merlin where and when the photo was taken. To orient Merlin, the user draws a box around the bird and clicks on its bill, eye, and tail. Merlin does the rest. Within seconds, it looks at the pixels and combines powerful artificial intelligence techniques with millions of data points from humans, then presents the most likely species, including photos and sounds.
"Computers can process images much more efficiently than humans — they can organize, index, and match vast constellations of visual information such as the colors of the feathers and shapes of the bill," says Serge Belongie, a professor of Computer Science at Cornell Tech and an author of the paper. "The state-of-the-art in computer vision is rapidly approaching that of human perception, and with a little help from the user, we can close the remaining gap and deliver a surprisingly accurate solution."
Other co-authors of the CVPR 2015 paper are Grant Van Horn, Steve Branson, Ryan Farrell, Scott Haber, Jessie Barry, Panos Ipeirotis, and Pietro Perona.
Merlin's success relies on collaboration between computers and humans. The computer learns to recognize each species from tens of thousands of images identified and labeled by bird enthusiasts. It also taps in to more than 70 million sightings recorded by birders in the eBird.org database, narrowing its search to the species found at the location and time of year when the photo was taken.
Because the photo identifier uses machine learning techniques, it has the potential to improve the more people use it. After it can reliably identify photos taken with smartphones, the team will add it to the Merlin Bird ID app, a free app that has helped users with more than one million bird identifications by asking them five questions.
Merlin's computer vision system was developed by Steve Branson and Grant Van Horn of the Visipedia project, led by professors Pietro Perona at the California Institute of Technology and Serge Belongie at Cornell Tech. Their work was made possible with support from Google, the Jacobs Technion-Cornell Institute at Cornell Tech, and the U.S. National Science Foundation.
No entries found