Laudando-Associates-LLC/pucks-ros2

<h1 align="center"><strong>Pucks ROS 2 dataset</strong></h1>

Dataset Overview

<div align="justify">

The Pucks Dataset is part of the L&Aser surrogate target recognition pipeline developed by Laudando & Associates LLC. This ROS 2 dataset contains high-resolution field images of circular surrogate targets (“pucks”) used to simulate real weed targets for the L&Aser weeding system.

</div>

Contents

• —The training, validation and test dataset used to train D-FINE.

• —Raw ROS 2 Bag file recorded from a robot.

Requirements

ROS 2 is mandotory to run the Bag file.

• —It can be installed by following the Offical Docs.

• —Using a ROS 2 Docker container from the Docker Hub.

• —Using a conda package manager from RoboStack.

Setup

Before downloading the dataset, install the huggingface_hub Python library:

pip install huggingface_hub[hf_xet]

How to Download

from huggingface_hub import snapshot_download
import os

repo_id = "Laudando-Associates-LLC/pucks-ros2"
target_folder = os.path.join(os.getcwd(), "04-23-2024_4w")
files_to_download = ["metadata.yaml", "test_0.db3"]

os.makedirs(target_folder, exist_ok=True)

for filename in files_to_download:
    snapshot_download(
        repo_id=repo_id,
        repo_type="dataset",
        local_dir=target_folder,
        allow_patterns=[filename]
    )
    print(f"Downloaded '{filename}' to '{target_folder}'")

The test_0.db3 is around ~20Gb.

How to use

In your ROS 2 enviroment

ros2 bag play <path>/<to>/04-23-2024_4w --clock

Output Topics

Recording Date

<div align="justify">

The data was originally collected in ROS 2 bag format in April 2024 during controlled field trials conducted by Laudando & Associates LLC.

</div>

License

<div align="justify">

This dataset is provided under the L&Aser Dataset Replication License (Version 1.0). See LICENSE for full terms.

</div>