07-13, 14:35–15:05 (Europe/Dublin), Liffey Hall 1
There are over 70 000 different Lego bricks and they appear in almost 200 different colors. Even the most hardcore AFOLs (Adult Fan of Lego) don’t know all of them. Let alone be able to recognize them. So I got curious whether it’s possible to create an application that can recognize the particular brick using only its photo.
During this talk, I will walk you through my journey to create the Lego bricks recognition application in Python. I will start with dataset creation and introduce some Lego-specific concepts and resources. Then, I will explain a few different Machine Learning approaches to solve Lego bricks recognition task – classification, detection, and metric learning. Finally, I will show what has worked, what hasn’t, and how you can play with it yourself.
some
Expected audience expertise: Python –some
Abstract as a tweet –Building an app that recognizes 70 000 different Lego bricks!
Piotr Rybak is a Machine Learning Researcher with experience in industry and academia. In his work, he mainly focuses on Natural Language Understanding but once in a while, he likes to dive into other topics. In his free time, he's a big fan of board games, Lego bricks, and boulder climbing.