genBoxes

This program generates synthetic 3D point clouds, sampled from different simple object types (boxes, simplified cars (2 boxes side-by-side), spheres, cylinders) and with custom sampling parameters. The goal is to simulate simplified 3D Lidar scenes, aimed more towards autonomous driving scenes.

Tip

The generated output can be used directly in LearnableEarthParser through EarthParserDataset, and more specifically in the corresponding forks specifically dedicated to it: LearnableEarthParser and EarthParserDataset.

Set up environment and activate it

conda env create -f genboxes.yml
conda activate genboxes

How to use

Without visualisation

python main.py

With visualisation

python main.py settings.visu=True

Configuration options

Custom generation configurations can be created, using Hydra as a backend. To understand how it works, you can have a look at the configs folder containing:

• a default.yaml file with the default configuration;
• and all the other config files that already contain advanced settings.

You can launch them by using the following command:

python main.py --config-name config_name