MaCAD 2025 / Master Thesis - Name TBD

This Master Thesis project is a copilot to improve architects knowledge on a site to improve masterplanning, by leveraging OSMX, GraphML (2D & 3D), LLM, UI, ...

1_Brief to Graph

Brief to Graph step is translating natural language, not detailed, into a pair of csv for nodes and edges, enabling a graph representation, by using Large Language Model (LLM) (model to be determined e.g. qwen3:8b or llama3.1:8b) and Natural Language Preprocessing (NLP) preprocessing.

Input - Masterplan brief:

• e.g., “A mixed-use neighborhood along the riverfront... high walkability... civic plaza... integration with existing transport lines...”

Output - CSV of nodes & edges:

• CSV of nodes: fields like:
  • id, program, scale or size, denisty, typology, social weight or value, mobility relevance, ...
• CSV of edges:
  • source_id, target_id, connection_type, intensity, integration_with_city, ...

In between steps

• brief preprocessing with NLP
• brief to json with LLM
• json to csv with LLM
• graph visualisation with Rhino/Gh

2_Graph to 2D layout

3_Graph to 3D layout

Authors

• César Diego Herbosa @cdherbosa
• Aymeric Brouez @Aymrc

Supervisor

• David Andrés León - IAAC MaCAD Director