This open dataset supports individual tree species identification in agroforestry settings using multimodal remote-sensing and tree-level metadata. Each example corresponds to a single tree and links across SkySat RGB crops, Sentinel-1 time series, Sentinel-2 time series, PlanetScope time series, species labels, and derived species-classification vectors. The release contains 23,156 tree examples. Raw Google Street View imagery is not redistributed; panorama IDs and related metadata are retained to support reproducible workflows where users have appropriate API access and rights. Manifests and metadata files are included to support reliable joins across modalities and data splits.
# Species Identification Dataset Context This dataset supports individual tree species identification from multimodal remote-sensing data and structured tree metadata. Each example corresponds to an individual tree and links across several released data views: - SkySat RGB crop imagery. - Sentinel-1 pixel time series. - Sentinel-2 pixel time series. - PlanetScope pixel time series. - Tree-level metadata, labels, and derived species-classification vectors. Raw Google Street View images are not included in this release. The release retains GSV panorama IDs and related capture/geometry metadata, and code is released separately to repopulate GSV imagery from those panorama IDs where users have appropriate API access and rights. The manifests document how examples align across modalities. Users should rely on the manifest files and `tree_metadata.csv` rather than inferring joins only from filenames.