Research

Interested in combination of CV / CG / NLP.

• UoE Bachelor Dissertation: Inverse Procedural Modeling: from Sketches to Buildings
  • Construct a procedural model for buildings based on Directed Acyclic Graphs (DAG)
  • Distort 3D models and render as a 2D sketch image; generate training data with parameter sampling
  • Encoder-decoder & Multi-task decoders that predict DAG parameters based on input sketch image
  • Develop Blender add-on as user interface
• GraphviX Lab, IPAB, UoE - Summer Research Internship (Summer 2023)
  • Researched on topics of sketch-based inverse procedural modeling. Implemented shape grammars based on Blender geometry nodes and its APIs. Trained neural networks to inference shape parameters from sketches, and integrated above pipelines as a Blender plug-in. Some tech stacks are also part of my bachelor dissertation.

Proposals

• Novel Procedural Shape Program Synthesis full / short