Remember that big box of Lego you had as a child that was filled with hundreds of different styles of Lego and twenty Lego men and women from Darth Vader to scuba divers? Trying to sort those boxes would have been impossible. That's probably what Daniel West experienced before he built his Universal Lego Sorting Machine from Lego. It uses artificial intelligence to sort the Lego pieces and we all want one for our ahem nieces and nephews.
Want to read some more interesting articles? Have a look at this story about the very strange fish that washed up on a beach in California.
This article was originally reported by The Verge.
After two years in the making, the Universal Lego Sorting Machine is completed and we think it's very cool. The machine is built with over 10,000 Lego pieces, and it's powered by six Lego motors and nine servo motors.
The machine can recognize and Lego part that's ever been made - even ones it hasn't seen before. It sorts them into 18 buckets at a rate of one brick every two seconds. West was inspired by two Lego sorting machines that have been built in the past, but his is the next step in the evolution of Lego sorting machines.
But how does the machine sort the Lego bricks - even ones it's never seen before? The type of artificial intelligence West used is called a convolutional neural network. It learns how to recognize Lego parts by training on realistic 3D images, meaning that it can develop knowledge about any Lego part for which there is a 3D model available. This is the first time something like this has been built.
The machine has three main components that solve different problems. The first problem is separating one piece of Lego from the huge bucket of parts that has been poured in the machine to show the camera for sorting. West solved this by creating two belts that slowly feeds Lego pieces onto the vibration feeder, which shakes the bricks around and separates them from each other.
The Lego pieces then fall into the scanner, which slowly moves the bricks past the camera, underneath a light. This footage gets processed by the Raspberry Pi computer, then this information is sent wirelessly to a more powerful computer that is able to run the neural network to classify the parts. The result of this classification gets sent back to the sorting machine and the Lego piece is sent to the correct output bucket.
There you have it. A machine that could have saved us hours of sorting (or not) when we were kids. Maybe this kind of machine will be available to buy in the future, saving the next generations the trouble of rummaging through a box of Lego to find the right piece. How different their lives will be.