How do birds get their colors? A guide to your vibrantly hued, winged friends by USA Today

For this project, we analyzed over 500 photographs of bird species provided by the Macaulay Library at the Cornell Lab of Ornithology, whose extensive media collection helps researchers identify and study color abnormalities and rare bird colors. The species were chosen based on data from Project FeederWatch, a citizen science program run by the Cornell Lab, which allows people to submit sightings of birds they've seen locally.

Using a machine learning algorithm, USA TODAY determined the dominant color of each bird photo. Photos were manually selected to find images of birds in a similar pose, with good lighting and for overall quality. The background was removed from all photos and each photo was run through the K-means algorithm to get the dominant color. K-means works by dividing the colors of a bird into distinct groups, or "clusters" to find the most representative colors. Each pixel's color is then assigned to the nearest cluster center. The cluster with the most pixels becomes the dominant color.

We then combined this technical approach with explanatory and interactive graphics.