May 2018 | By Gaël Hugo
Browse through the 20th century via Tags defined by Machine Learning.
LIFE Tags organizes over 4 million images from the Life magazine archives into an interactive encyclopedia using machine learning. Running between 1936 and 1972, Life magazine was an American news publication with an emphasis on photojournalism. Our experiment brings the iconic magazine to life. Browse, search and rate images corresponding to thousands of labels as funny, right or wrong.
With machine learning, we classified millions of images automatically into a catalogue format based on thousands of labels. For this project, we used Google’s Image Content-based Annotation (ICA) algorithm to generate labels based on image pixels. It is based on a deep neural network used in Google photo search that has been trained on millions of images and labels to recognize categories for labels and pictures.
You can easily navigate the Life magazine archive by browsing image categories and labels. We clustered labels into categories using the nearest neighbor algorithm, which finds related labels based on image feature vectors. The ICA algorithm provides digital information for each image, which we translated into keywords, compared and grouped together.
Each image has multiple labels linked to the elements that are recognized. Click to view the full-size image and the dotted lines will show you where the relevant elements are located.