Keypoint Autoencoders

This repository contains code of the paper Keypoint Autoencoders: Learning Interest Points of Semantics.

Training

To run the training process:

import acae  # Choose one from `acae` and `vnapf`
acae.train()  # Trains the model
acae.visual_test(True)  # Picks a point cloud in the test set and visualize the results

Data Preparation

The data should be stored in the ./point_cloud/train and ./point_cloud/test.

All .h5 files under those folders are loaded. Each file should contain a data array of shape (n, m, 3) which is n point clouds with m points, and a label array of shape (n) which indicates the classes of the point cloud. m is required to be same for all point clouds. The paper uses point clouds generated from the ModelNet40 dataset, which can be downloaded here.