Automated LEGO Sorting System

The problem

Collectors, resellers and educators handle thousands of loose bricks. Sorting them by part and colour by hand takes hours and mistakes are common.

How it works

A hopper and vibration feeder separate bricks so they pass the camera one at a time. An OpenCV pipeline running in C++ on an embedded board segments each brick, a classifier identifies its part number and colour, and an air jet pushes it into the right bin.

Every sorted piece is logged to a cloud inventory. Operators follow throughput and bin levels live in a web dashboard and review low-confidence classifications, which feed the next training round.

System components

• Mechanical handling: hopper, vibration feeder, conveyor and ejectors driven by the controller.
• Edge vision: real-time capture, background subtraction, segmentation and inference in C++.
• Model training: dataset building and training in Python, exported to the edge runtime.
• Backend and data: sort-run API, parts catalogue and inventory in PostgreSQL, images in object storage.
• Operator dashboard: live throughput, bin levels, review queue and inventory export.
• Infrastructure: containerised services, infrastructure as code and CI/CD for edge and cloud software.