RobotCarDataset Scraper

Introduction

This script can be used to perform sequenced downloads of various sensors logs for the robotcar-dataset https://robotcar-dataset.robots.ox.ac.uk/.

The code is tested for Python3 on Ubuntu 16.04.

Getting started

A prebuilt docker image is available with:

docker pull matthewgadd/robotcar-dataset-scraper:latest

Otherwise, you can build the docker image from scratch:

wget https://raw.githubusercontent.com/matthewgadd/RobotCarDataset-Scraper/master/Dockerfile
docker build -t matthewgadd/robotcar-dataset-scraper:latest .

Example usage

Mount your downloads dir through the docker image:

docker run --rm -it -w /RobotCarDataset-Scraper -v $HOME/Downloads:/Downloads matthewgadd/robotcar-dataset-scraper:latest

The script

python get_datasets.py

parses the html for the complete dataset listing and scrapes each dataset page and retrieves the available files (tar) for download. With the example output file datasets.csv provided, you should then run:

python scrape_mrgdatashare.py --downloads_dir /Downloads --datasets_file datasets.csv --username USERNAME --password PASSWORD

If you do not need the full dataset, you can use --choice_sensors and --choice_runs_file options to download parts of the dataset on your own needs.

• --choice_sensors option can receive multiple sensor names in tags, stereo_centre, stereo_left, stereo_right, vo, mono_left, mono_right, mono_rear, lms_front, lms_rear, ldmrs, gps, all.
• --choice_runs_file option receive a .txt file that contains the names of runs you want to download, we provide an example filea sample file example_list.txt .

for example you can download "stereo_centre", "vo" and "lms_front" data of "2014-05-19-13-20-57" and "2014-06-26-09-31-18" by the following command:

python scrape_mrgdatashare.py --choice_sensors stereo_centre,vo,lms_front --choice_runs_file example_list.txt --downloads_dir /Downloads --datasets_file datasets.csv --username USERNAME --password PASSWORD