africa-field-boundary-labels.yaml

Name: A region-wide, multi-year set of crop field boundary labels for Africa
Description: >
  Crop field boundaries digitized in Planet imagery collected across Africa
  between 2017 and 2023, developed by [Farmerline](https://farmerline.co/), [Spatial Collective](https://spatialcollective.com/), 
  and the [Agricultural Impacts Research Group](https://agroimpacts.info/) at [Clark University](https://www.clarku.edu/), with support from the
  [Lacuna Fund](https://lacunafund.org/) ([Estes et al, 2024](https://arxiv.org/abs/2412.18483); [Wussah et al. (2023)](https://zenodo.org/records/11060871)). This dataset has been 
  further supplemented by additional labels collected primarily for 
  for 2018 over a subset of countries, which provide an example of their
  application in training and validating a CNN-based cropland mapping model 
  [(Khallaghi et al. 2025)](https://www.mdpi.com/2072-4292/17/3/474).
Documentation: Information on the primary dataset can be found [here](https://github.com/agroimpacts/lacunalabels/).
  Documentation for added labels is available [here](https://github.com/agroimpacts/cnn-generalization-enhancement).
Contact: airg@clarku.edu
ManagedBy: "[The Agricultural Impacts Research Group](https://agroimpacts.info/)"
UpdateFrequency: "Updated versions of the dataset are added as they are developed"
Tags:
  - agriculture
  - machine learning
  - land cover
  - satellite imagery
  - cog
  - labeled
License: "[Planet NICFI participant license agreement](https://assets.planet.com/docs/Planet_ParticipantLicenseAgreement_NICFI.pdf)"
Resources:
  - Description: Field boundaries and corresponding Planet images
    ARN: arn:aws:s3:::africa-field-boundary-labels
    Region: us-west-2
    Type: S3 Bucket
  - Description: '[Additional rasterized field labels and corresponding Planet images](https://www.mdpi.com/2072-4292/17/3/474)'
    ARN: arn:aws:s3:::africa-field-boundary-labels/extra
    Region: us-west-2
    Type: S3 Bucket
DataAtWork:
  Tutorials:
    - Title: Instructions on data access and label-making demonstration notebook
      URL: https://github.com/agroimpacts/lacunalabels
      NotebookURL: https://github.com/agroimpacts/lacunalabels/blob/main/notebooks/makelabels/label-chips.ipynb
      AuthorName: Lyndon Estes
      AuthorURL: https://github.com/ldemaz
  Publications:
    - Title: Generalization enhancement strategies to enable cross-year cropland mapping with convolutional neural networks trained using historical samples
      URL: https://www.mdpi.com/2072-4292/17/3/474 
      AuthorName: Khallaghi et al. (2025)
    - Title: A region-wide, multi-year set of crop field boundary labels for Africa
      URL: https://arxiv.org/abs/2412.18483
      AuthorName: Estes et al. (2024)
    - Title: A region-wide, multi-year set of crop field boundary labels for Africa
      URL: https://zenodo.org/records/11060871
      AuthorName: Wussah et al. (2023)
    - Title: Technical report on label develop and processing
      URL: https://github.com/agroimpacts/lacunalabels/blob/main/docs/report/technical-report.pdf
      AuthorName: Wussah et al. (2023)
    - Title: High resolution, annual maps of field boundaries for smallholder-dominated croplands at national scales
      URL: https://www.frontiersin.org/article/10.3389/frai.2021.744863
      AuthorName: Estes et al. (2022)
    - Title: A platform for crowdsourcing the creation of representative, accurate landcover maps
      URL: http://www.sciencedirect.com/science/article/pii/S136481521630010X
      AuthorName: Estes et al. (2016)
Citation: >
  Primary dataset: Estes et al. (2024). A region-wide, multi-year set of crop field boundary labels for Africa. arXiv:2412.18483.
  Additional labels: Khallaghi et al. (2025). Generalization enhancement strategies to enable cross-year cropland mapping with convolutional neural networks trained using historical samples. Remote Sensing, 17(3), 474.