Point cloud tutorial
This repository contains tutorial code and supplementary note for point cloud processing. In order to understand algorithm of point cloud processing, most codes are implemented with Numpy and Jupyter.
Repository structure
┌─ data # 3D files for examples
├─ .devcontainer # Dockerfile for this tutorial
└─ python # python codes of tutorial
├─ tutlibs # package for tutorial codes
└─ *.ipynb # tutorial codesHow to use
1. Enviroment
You can execute most tutorial codes on Codespaces(CPU resource). If you have GPU resource, can execute all tutorial codes. For more environment of CPU and GPU, Please refer to .devcontainer/README.md.
2. About tutorial
Tutorial contents are as follows:
Basic
| Theme | Page & code | Contents | Other packages list | Todo |
|---|---|---|---|---|
| Basic code | python | We introduce basic code used in this tutorial. | ||
| Characteristic | python | characteristic of the point cloud | ||
| Nearest neighbors search | python | kNN, Radius Search, Hybrid Search | python, C++ | |
| Tree structure | python | None | ||
| Downsampling | python | Random Sampling, Furthest point sampling, Voxel grid sampling | ||
| Normal estimation | python | Estimation with PCA, Normal re-orientation methods | ||
| Handcrafted feature | python | FPH, PPF | ||
| Visualization | python | Visualization functions in the repository | ||
| Competition | python | None | add description | |
| Transformation/Affine transformation | python | Affine transformation, Transformation matrix | ||
| Transformation/Camera projection | python | Camera projection | ||
| Task/Reconstruction | python | Marching cube, meshes to points | add description | |
| Task/Registration | python | ICP, RANSAC with Handcrafted features | add description | |
| Task/SLAM | python | None | python |
Deep Learning
| Theme | Page & code | Paper name | Other packages list | Todo |
|---|---|---|---|---|
| PointNet | python | PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation | translation to english | |
| PointNet++ | python | PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space | translation to english | |
| VoxNet | python | VoxNet: A 3D Convolutional Neural Network for Real-Time Object Recongnition | translation to english |
Dataset
| Dataset Name | Page & code | Paper name | Other packages list | Todo |
|---|---|---|---|---|
| Pix3D | python | Pix3D: Dataset and Methods for Single-Image 3D Shape Modeling | ||
| Redwood 3DScan | python | A Large Dataset of Object Scans | ||
| Redwood Indoor | python | Robust Reconstruction of Indoor Scenes | ||
| ScanNet | python | ScanNet: Richly-annotated 3D Reconstructions of Indoor Scenes | ||
| Sun3D | python | SUN3D database |
About correction
If you finded any corrections, let us know in Issues.