Bricks of the Trade: Techie Harnesses AI Skills to Sort Son’s Legos

by Brian Caulfield

Francisco “Paco” Garcia is a dad. And being a dad these days means you have Legos. Way too many Legos.

But Paco’s no ordinary dad. This is a dad with a GPU, a GPU he put to use doing what every dad dreams of doing: sorting those Legos out once and for all.

Our podcast’s host, Noah Kravitz, another dad with too many Legos, tracked Garcia down and asked him to share his secrets.

Garcia shared with us how — inspired by a Japanese cucumber farmer — he designed, built, and documented his work building a Lego sorter using TensorFlow, Raspberry Pi, and his NVIDIA GPU.

“You have this small farmer with cucumbers assembling, in a do-it-yourself manner, an intelligent cucumber sorter,” said Garcia, whose day job is in the technology industry.

“It’s still something that intrigues me, how his this going to work in the future, from self driving cars, to all these other types of use cases out in the world,” Garcia said.

Inspired by the idea that deep learning is something that can tackle small problems — as well as the big stuff — he looked to apply what he’d learned to the Lego sets scattered around his house.

It was a journey Garcia’s shared in a series of blog posts documenting his project that have been widely read. One of his guiding principles: keep it simple.

“I really wanted to see how good a neural network could be out of the box,” Garcia said. The tough part: the mechanical stuff. The result — set up on Garcia’s family room table — is a wonder of home brewed automation able to sort 11 types of Lego.

To hear the whole story, tune in to the latest edition of our AI podcast.

How to Tune in to the AI Podcast

Our AI Podcast is available through iTunesCastbox, DoggCatcher, Google Play MusicOvercastPlayerFMPodbay, Pocket Casts, PodCruncher, PodKicker, Stitcher and Soundcloud. If your favorite isn’t listed here, email us at aipodcast [at] nvidia [dot] com.