YIBADA

Cornell Website Uses Image Recognition Algorithm To ID Bird Species

| Jun 07, 2015 08:36 AM EDT

Merlin Bird Photo ID

Cornell University's new website features its artificial intelligence (AI) software that can identify bird species in the United States and Canada. The image recognition algorithm works by searching its huge database of bird photography after users click body parts of the feathered animal, and input key data.   

The Cornell AI website is named Merlin Bird Photo ID. It provides bird species identification after users draw a box around the avian flying creature; and click on the eyes, beak, and tail.

Users must also input other data about the photographed egg-laying beings, including Canadian birds. That includes the animal's base color and its action when sighted, according to Tech Times.  

The picture recognition app then makes a match. It sifts through tons of pictures of the winged animals, including 3 million taken by amateur birdwatchers and 70 million images from the eBird.org database.   

Then within seconds the Cornell website will show the best matches, such as for American birds. It also shows photos and plays songs of the various birds species, according to Engadget.  

The photo recognition software's identification skills of various species improve with more repetitions. That is like other learning machines.  

Serge Belongie, Cornell Computer Science professor, explained that computers process pictures more efficiently than people can. The user helps the AI software to almost mirror human perception.

In fact, Jessie Barry, the Merlin Project Leader, shared that the  online application has a "truly amazing" 90 percent success rate including the correct bird type in the top three results. This is after just a few years of development.   

Merlin Bird Photo ID only works on PCs and laptops. However, a bird species ID app will soon will be available for Android and iOS smartphone platforms.

Related News

Most Popular

EDITOR'S PICK