OPTIMIZED SYSTEMS

Image Search Tool Estimates Human Age From Photos

Rebecca Kutzer-Rice | Date: 12-30-11 | Comments

Using an automatic search tool, researchers have created the largest database of human faces with age-related data. With the database, a computer can now estimate a person�s age with remarkable accuracy.

Although today�s computers can rapidly identify objects and information, analyzing more subtle qualities�such as the age of a person in a photograph�proves difficult. People are much better than machines at guessing ages. But a new database of human ages might change that fact.

Bingbing Ni of the Advanced Digital Sciences Center in Singapore, in collaboration with Zheng Song and Shuicheng Yan from the National University of Singapore, developed the database with an automatic search-engine tool. The tool compiled more than 400,000 images from common Websites like Google and Flickr. It also automatically collected 10,000 video clips from YouTube.

The new database, compiled automatically, contains hundreds of thousands of human faces and age-related information.

After collecting the images, the system automatically tagged them with the ages found on the Websites. Researchers then removed poor-quality and incorrect images. Ultimately, they were left with a database with more than 219,000 faces.

The new database has far more data than any other similar information bank. It also includes faces from different racial groups and images with different lighting conditions. This gives it an ability to analyze a broad range of faces from a wide range of photographs.

�The greatest significance is that, using Web-based mining, there is no need to manually construct a face image database, which is quite labor-intensive and costly,� Ni told PhysOrg.com. �And the automatically mined Web image database possesses the generalization capability, e.g., models trained on this data can be applied to general faces.�

Using the database as a model, the computer tool can estimate ages with accuracy to within five years.

Researchers hope that the age estimator could be used in a variety of ways. The tool could analyze security footage to guess the ages of suspects. It could also help collect demographic data and improve human identification.

�For example, digital signage applications�installed in shopping malls, vending machines, etc., automatically estimate the customers� ages and recommend corresponding product advertisements,� Ni said.

The team�s results, �Web Image and Video Mining toward Universal and Robust Age Estimator,� were recently published in "IEEE Transactions on Multimedia."