Birder's Guide

NOV 2017

Birder's Guide is the American Birding Association's newest publication. Each issue focuses on a key subject, providing tips from experienced birders on a wide variety of topics like Travel, Listing & Taxonomy, Gear, and Conservation & Community.

Issue link:

Contents of this Issue


Page 20 of 51

Challis, Idaho Challis, Idaho around them. At the Cornell Lab of Ornithol- ogy, we develop tools aimed to transform the way in which people of all skill levels inter- act with birds. The best part is that the free Merlin Bird ID app is possible because of you: your contributions of sightings, photos, and sounds to eBird and the Macaulay Library. With eBird data, Merlin is able to synthesize data from experienced birders to help new birders get started. That said, Merlin isn't just for beginners. New features and Merlin's regional expan- sion are increasing the utility of this app for experienced birders, too. What Makes Merlin Tick: eBird When we were first brain- storming the idea of creat - ing a bird app at the Cornell Lab, a lot of time and effort was spent thinking about how expert birders identify birds. Sure, what a bird looks like is very impor- tant to identifying it, but equally important is understanding status and distribution—when and where a particular bird is expected to be found—something that takes years of intense birding to grasp. With birders submitting mil- lions of sightings to eBird, it is now possible to create an app that can simulate what an ex- pert birder in any region intuitively knows— which species of birds are likely to occur at a given date and location. Likelihood of occur- rence is arguably the biggest deciding factor for most bird identifications. In Merlin's step-by-step identification fea- ture (found via the Start Bird ID button), status and distribution information plays a key role; the first questions Merlin asks are about the date and location for the bird you are identifying. With the provided date and location, Merlin pulls in the species reported nearby by your fellow birders. This list of ex- pected species is combined with information on color, size, and behavior to return a short list of birds for you to consider. Automatic ID of Bird Photos Ten years ago, who could have imagined that your phone could identify peeps, sparrows, and gulls? Com- puter vision and machine learning technology are be- coming part of our everyday lives, and these technologies are becoming part of bird- ing. Merlin has built-in com- puter vision algorithms that offer identification help for more than 1,500 species. Give Merlin a photo, and in just a couple seconds—without an internet connection— Merlin will give you an answer! In 2012, the Merlin team began a collabo- ration with the Visipedia team—computer vision researchers at Caltech and Cornell Tech developing algorithms to automatically analyze images and extract knowledge from the visual world—to take on the challenge of teaching computers to identify birds in im- ages. The first tests yielded only 18% accu- racy. Today, on the same dataset, accuracy is over 90%. These improvements come thanks to recent breakthroughs in the field of machine learn- 19 November 2017 | Birder's Guide to Gear Drew Weber Jessie Barry Ithaca, New York Ithaca, New York Merlin Bird ID

Articles in this issue

Links on this page

Archives of this issue

view archives of Birder's Guide - NOV 2017