Submission + - Man Sorts Two Metric Tons of LEGO Via Machine Learning 1
adeelarshad82 writes: Lego enthusiast, Jacques Mattheij, ended up winning bids for 4400 pounds on ebay. Mattheij had noticed that unsorted bricks sell for something like €10/kilogram, whereas sets are roughly €40/kg and rare parts go for up to €100/kg. Hence, much of the value of the bricks is in their sorting. If he could reduce the entropy of these bins of unsorted bricks, he could make a tidy profit. Unlike most other people who sort these by hand, Mattheij wrote a software that used a camera paired with a computer running a neural net-based image classifier to sort the pieces. At its core, the system took images from a webcam and fed them to a neural network to do the classification. Given the sheer volume of size, color and orientation, the neural net needed to be "trained" by showing it lots of images, and telling it what those images represent. Mattheij's breakthrough was allowing the machine to effectively train itself, with guidance: Running pieces through allows the system to take its own photos, make a guess, and build on that guess. Here's a video of the system in action.
