RedwoodNeRF

A Python toolbox for NeRF-accelerated forest assessment

Datasets

We provide two datasets totaling in 540 m² of coastal California mixed-evergreen Redwood forest. This study area belonds to the Forest Ecology Research Plot FERP, a globally recognized ForestGEO plot spanning 16 ha and over 51,000 recorded stem diameters. All point clouds are in world scale, but the NeRF recostructions must undergo a transformation provided in an associated file called dataparser_transforms.json.

• Dataset A (140 m²) Google Drive
• Dataset B (400 m²) Google Drive

Tree segmenation and modeling (coming soon...)

Our software tools are built on top of TreeTool, a Python library for estimating tree diameters from dense forest reconstructions. We provide a full add-on library which expands TreeTool's viability for irregular tree shapes and complex tree growth patterns. We also introduce an improved tree modeling approach which uses stacks of convex-hull slices to accuratley represent irregular tree morphologies.

Paper

This effort is based on our paper which is currently under review. Please cite our preprint if you use the provided code or datasets.

@article{koryckiRedwoodNeRF,
      title={NeRF-Accelerated Ecological Monitoring in Mixed-Evergreen Redwood Forest}, 
      author={Adam Korycki and Cory Yeaton and Gregory S. Gilbert and Colleen Josephson and Steve McGuire},
      year={2024},
      eprint={2410.07418},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2410.07418}, 
}

Please also cite TreeTool.

@article{montoya2021treetool,
  title={TreeTool: A tool for detecting trees and estimating their DBH using forest point clouds},
  author={Montoya, Omar and Icasio-Hernandez, Octavio and Salas, Joaquin},
  journal={SoftwareX},
  volume={16},
  pages={100889},
  year={2021},
  publisher={Elsevier}
}