ReNet Library

This project is consist of 2 parts

• ReNet
• EnvioX

ReNet is Reconstruction Network which is specialized for reconstructing spatial information from noisy, blinded sensor measured data. For more detailed information, please read the paper.

EnvioX is library that helps to generate and process 3D point cloud & voxel data for training ReNet.

Main framework of this project is MinkowskiEngine. Please read the document for more information.

Features

• Custom environment point cloud data (with boxes, pillars, and walls)

• terrain point clould & voxel data around a robot position

  

• terrain point cloud & voxel data detected by sansors on a robot (noisy, blinded)

  

• point cloud & voxel data Visualizer

• ReNet training environment including ReNetDataset, ReNet_collate_fn

Requirments

Please refer to MinkowskiEngine Requirements.

How to Use

Install MinkowskiEngine following the installation guide.

Next, install matplotlib & scipy & numpy

pip install matplotlib scipy numpy

Then, download ReNet & EnvioX to your project directory, and import desired features.

import sys
from pathlib import Path
sys.path.append(str(Path(__file__).parents[1]))

import EnvioX
import ReNet

Please refer to Example for more detail about how to use specific functions.

For Colab user

Refer to ReNet.ipynb. This notebook file contains from setting environments for MinkowskiEngine to data generation, train, validation and test.