Waymo Open Dataset Preparation

Adapted from Waymo-Dataset-Tool

This tool downloads Waymo Open Dataset and converts point clouds, annotations and poses into Argoverse2 format.

Installation

f you installed the conda environment from SeMoLi, all libraries are already installed and you can run the code if you activate conda activate SeMoLi. Otherwise, perpare a conda evironment running the following:

conda create -n waymo_od python=3.9
conda activivate waymo_od

Install gsutil (command line tool to download files from google cloud. Follow this Guide to install. After installing the Cloud SDK, simply run gcloud init, then you will be asked to log into your google account. Make sure it is the same account that has Waymo Dataset access.)

Finally, install waymo_open_dataset by running:

pip install waymo-open-dataset-tf-2-11-0==1.6.1

How to use

This tool supports downloading train, val and test set of waymo open dataset and extracts only the LiDAR information from the tfrecord files.

Download and extract train and val

To download and extract train and val sets run:

bash download_and_extract_all.sh

Download and extract single splits

To download and extract single splits run:

python batch_download.py <split> --out-dir data/Waymo
python convert_tfrecord.py <split> --record_path data/Waymo --out_dir data/Waymo_Converted

where again split is either train val, or test.