↪︎ 3D Point Cloud Classification with Geometric Deep Learning
This repository contains code for conducting experiments on 3D point cloud classification using various geometric deep learning architectures. The experiments are performed on the ModelNet40 dataset, which consists of 3D models from 40 different categories. The starting point for this repository was the DeepSets architecture for permutation invariant learning on sets (for example point clouds). We wanted to test wether using graph neural networks (GNNs) and more complex architectures could improve performance on this task by incorporating additional geometric priors.
The results were negative in the sense that none of the more complex architectures were able to outperform the simple DeepSets architecture. However, this is still a useful finding as it suggests that for this particular task, the added complexity of using GNNs may not be necessary. Furthermore, I believe the implementations of the different architectures and the dataset construction methods could be useful for future experiments and research in this area.
