Variational PointNet Encoder-Decoder for 3D Point Cloud Data

Welcome to the Variational PointNet Encoder-Decoder repository! This project represents my endeavor to create a solution for processing and generating 3D point cloud data. It offers a range of encoding and decoding options to suit various application needs.

Encoding Options

1. PointNet++

• The PointNet++ encoder is a robust and widely-used choice for extracting features from 3D point cloud data.

2. Convolutional Encoders

• My implementation includes convolutional encoders, designed to leverage spatial dependencies within the point cloud.

3. Graph Neural Network (GNN) Encoding (Future Plan)

• Stay tuned for upcoming GNN encoding capabilities, which will further enhance your 3D data processing options.

Decoding Options

1. Deconvoluting Decoder

• The deconvoluting decoder is a powerful tool for reconstructing 3D point clouds from latent representations.

2. Multi-Layer Perceptron (MLP) Decoder

• My repository includes an MLP decoder, offering flexibility and efficiency in point cloud generation.

3. Transformer-based Autoregressive Decoder (Future Plan)

• I am actively working on a state-of-the-art transformer-based autoregressive decoder for high-quality point cloud generation. Stay tuned for updates!

Get Started

To get started with my advanced Variational PointNet Encoder-Decoder, please refer to the documentation and instructions in the Wiki section. (under development)

Contribution and Support

For any issues or questions, feel free to open an Issue