Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
December 8, 2025ACM Transactions on GraphicsOpen Access

LEGO®-Maker: Autoregressive Image-Conditioned LEGO® Model Creation

View Full Paper
Ask AI
Bookmark
Share

Authors

JGJiahao GeMZMingjun ZhouHZHanyou Zheng

Discussion

Loading...

Member takes

Overview

Generative model produces detailed LEGO models from images, highlighting strengths in geometry and structural integrity.

Key Points

  • This work aims to develop LEGO-Maker, a model for generating LEGO structures based on image inputs.
  • Introduced a compact LEGO tokenization scheme for model serialization and autoregressive learning.
  • Developed an autoregressive architecture with multi-token prediction to manage multiple brick attributes.
  • Established a data preparation pipeline using a procedural generator and a LEGO-to-real image translator.
  • Demonstrated superior performance in generating LEGO models across evaluated categories.
  • Achieved high evaluation scores in geometry, color, semantics, and structural integrity compared to baselines.

Cite This Study

Ge et al. (2025) studied this question.

synapsesocial.com/papers/693624ce4fa91c937236ceb9https://doi.org/10.1145/3763285
View Full Paper
Ask AI
Bookmark
Share

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

  1. 1Creating LEGO Figurines from Single Images2024 · 2 citations
  2. 2TreeSBA: Tree-Transformer for Self-Supervised Sequential Brick Assembly2024
  3. 3LEGO Co-builder: Exploring Fine-Grained Vision-Language Modeling for Multimodal LEGO Assembly Assistants2025 · 1 citations
  4. 4A Lightweight and Transferable Design for Robust Lego Manipulation2024
  5. 5Fabrication-Aware Synthetic Dataset Generation and Compact Geometric Encoding for Architectural Robotic Assembly of Brick Wall Designs2025