Many of us grew up with LEGO and there is a good chance that many of us are still collecting LEGO even up till today. This means that by now, chances are you would have built up quite a collection of various LEGO pieces. Unless you’re the type of person that would organize their collection by sorting out the pieces by size, color, and/or type, chances are you probably have a couple of boxes with random pieces thrown in.
This means that when you want to build something new, it can be a pain to try and go through all of them. This is something that engineer Daniel West solved when he decided to build himself a LEGO sorting machine that relies on a neural network to identify, classify, and even organize all the pieces brick by brick.
The machine itself is built out of LEGO bricks, 10,000 or so pieces, and took West about two years to design and perfect it. It also contains a handful of LEGO motors and servos to control the conveyor belts which transports the pieces to a camera. The video from the camera is then captured and processed by a Raspberry Pi computer and then sent to a laptop that contains the neural network.
It basically matches the pieces on the belt to a massive database containing 3D models of every LEGO piece ever built, thus allowing the sorter to know what piece it is and where to send it.
It’s actually a pretty awesome idea and if you have a few minutes to spare, you can check out West’s creation in the video above.