{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "1da5bb76",
   "metadata": {},
   "source": [
    "# Welcome!  \n",
    "This is the Huggingface space provides a demo of the devkit tools for working with the Flooded Road Environments Dataset (FRED). This work aims to enable and foster more research into the improvement of methods for detecting water hazards from autonomous vehicles.  \n",
    "  \n",
    "Preprint Paper: https://arxiv.org/abs/2605.22018  \n",
    "Github Repo: https://github.com/AVR3-Training-Centre/python-FRED  \n",
    "Dataset Page: https://huggingface.co/datasets/CMalone-Jupiter/FRED  \n",
    "Research Group: https://avr3.org.au/ \n",
    "\n",
    "## Cite\n",
    "If you use this dataset or development kit in your work, you can cite our preprint article using:  \n",
    "```\n",
    "@article{malone2026fred,\n",
    "      title={FRED: A Multi-Modal Autonomous Driving Dataset for Flooded Road Environments}, \n",
    "      author={Connor Malone and Sebastien Demmel and Sebastien Glaser},\n",
    "      year={2026},\n",
    "      journal={Arxiv},\n",
    "      url={https://arxiv.org/abs/2605.22018},\n",
    "}\n",
    "```"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bddbf94b",
   "metadata": {},
   "source": [
    "## The Flooded Road Environments Dataset (FRED)  \n",
    "This autonomous vehicle dataset has been developed to enable research into the detection of flooded roads during on-road deployment. The dataset was collected using a Renault Zoe with custom modifications to enable autonomy, including front and rear Blackfly cameras, an Ouster OS1 LiDAR, and a GNSS-corrected IMU. Data has been collected using the vehicle's sensor stack from 5 separate locations around Brisbane, Australia, both during and after flooding events. Semantic labels are provided for images to enable the development of detection methods, and corresponding position information from the GNSS-corrected IMU has been provided across sequences to additionally enable localization research for these scenarios."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "65ee831e",
   "metadata": {},
   "source": [
    "## SDK Tools\n",
    "This development kit provides several tools including loading, visualisation, and evaluation tools. Currently this includes: \n",
    "\n",
    "- Projecting pointclouds onto images\n",
    "- Visualising semantic labels on images\n",
    "- Evaluating semantic segmentation predictions\n",
    "- Displaying corresponding images across sequences from the same locations\n",
    "- Plotting sequence trajectories\n",
    "- Evaluating Visual Place Recognition (VPR) performance\n",
    "- Creating pointcloud semantic labels from image annotations\n",
    "- Creating range images from pointclouds\n",
    "- Infilling missing ground points in pointclouds\n",
    "\n",
    "The SDK will continue to be developed and updated to improve the functionality and utility for the research community."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3d498384",
   "metadata": {},
   "source": [
    "## Install\n",
    "Currently the SDK comes with a pixi.toml file that can be used to set up a python environment using the pixi package manager. In addition, a DockerFile is provided for use in this Huggingface app for online usage of the SDK and dataset. This may be able to be used to create a Docker environment on your own computer, however, this functionality is not tested."
   ]
  }
 ],
 "metadata": {
  "language_info": {
   "name": "python"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}