Topology-Aware Learning for 3D Point Clouds
Bachelor's thesis on geometric and topological inductive biases in deep learning for 3D point clouds.
Bachelor’s thesis project at SUSTech, supervised by Prof. Yifei Zhu (Dec 2025 – May 2026).
- Analyzed geometric and topological inductive biases in deep learning frameworks, evaluating persistent homology for robust feature extraction in unstructured 3D vision data.
- Proposed a two-axis taxonomy based on representation mechanisms and integration strategies for topology-aware point cloud learning.
- Designed and executed empirical ablation studies using PointNet backends to quantify performance gains from multi-modal feature fusion (topological descriptors + geometric coordinates).
Download: Thesis PDF