This repository is a curated list of resources for applying machine learning to LEGO bricks, targeting hobbyists, researchers, and developers interested in LEGO-related AI projects. It provides a comprehensive overview of projects, papers, and datasets focused on LEGO part recognition, sorting, set generation, and assembly.

How It Works

The resources are categorized by application area, including parts classification, sorting machines, set generation, assembly, image generation, 3D reconstruction, and price prediction. Projects leverage various ML techniques such as deep learning (CNNs, Mask R-CNN, GANs, Stable Diffusion), computer vision, and reinforcement learning, often utilizing custom datasets or synthetic data generated from LDraw models.

Quick Start & Requirements

This is a curated list, not a runnable project. Specific project requirements vary widely. Links to individual project code, papers, and datasets are provided.

Highlighted Details

• Extensive coverage of LEGO brick recognition and sorting systems, from DIY projects to academic research.
• Numerous datasets and rendering tools for generating synthetic LEGO data for ML training.
• Projects exploring generative AI for creating LEGO sets, minifigures, and architectural sculptures.
• Resources on robotic LEGO assembly and disassembly, and predicting LEGO set values.

Maintenance & Community

The list is maintained by 360er0. It includes links to various project blogs, GitHub repositories, and academic papers, indicating an active community around LEGO and AI.

Licensing & Compatibility

Licensing varies by individual project. Users must consult the specific licenses of linked repositories and datasets.

Limitations & Caveats

Many listed projects are research prototypes, unreleased, or appear abandoned. The quality and availability of code and datasets vary significantly across the listed resources.