At some point, you've probably sat down and rifled through a 'Where's Waldo' book. Some of us technical types probably wondered if there could be an automated way to find Waldo.
A new robot named 'There's Waldo' has mastered the art of finding our red-and-white striped shirt and beanie hat, Waldo in just 4.5 seconds.
'There's Waldo' was created by Matt Reed, a creative technologist at RedPepper. Reed and his team developed the robot using an Arm Swift Pro which is controlled by a Raspberry Pi computer. On the frontal area of the arm, a Vision Camera Kit takes an image of the Waldo puzzle and scans it through the computer vision program called OpenCV. Its goal is to recognize faces in the crowd as illustrated in a puzzle with Waldo in it.
'There's Waldo' has taken on the challenging task of finally finding Waldo.
The secret to this lies in Google's AutoML Vision - an easy-to-use computer system that allows users to train AI without any previous coding experience. Reed used the help of 62 Waldo facial images to train his AI which isn't a lot of data to use considering what AI usually runs on, but it works well.
In a video, the 'There's Waldo' robot is shown solving the puzzles in just a few seconds. When the robot first starts to search for Waldo, it takes a photo, picks out the faces that closely resemble Waldo based on anything higher than 95% confidence. After it has picked out the face that closely resembles Waldo, the robot uses its tiny hand to select the best option. However, its timing for selecting a Waldo character depends on the number of characters on the page and can take between 4 and 30 seconds to make a selection.
The creation of the 'There's Waldo' robot was inspired by Reed after seeing the Amazon Rekognition for celebrities. It only took Reed a week to develop the program which is impressive because some of us are still trying to find Waldo...with no luck still.
Have a story tip? Message me at: cabe(at)element14(dot)com
http://twitter.com/Cabe_Atwell
Top Comments