Defect GNN Visualizer
Explore crystal structures and their graph representations for predicting vacancy formation energies in perovskite materials. This tool visualizes the data pipeline that transforms atomic coordinates into inputs for Graph Neural Networks.
About This Project
This is a C++ implementation of the defect formation energy prediction pipeline, compiled to WebAssembly for browser-based visualization. The approach combines graph neural networks with persistent homology features to achieve state-of-the-art accuracy.
Note: The GNN training module is currently under development. This visualizer demonstrates the data preprocessing stages: structure parsing, neighbor list construction with Periodic Boundary Conditions, and graph assembly.