yxyyeah/PlantTraitsPrediction

Plant Traits prediction (from a Kaggle competition) using Vision Transformer and Autogluon Tabular Predictors.

Reference:
Schiller, C., Schmidtlein, S., Boonman, C., Moreno-Martínez, A., & Kattenborn, T. (2021). Deep learning and citizen science enable automated plant trait predictions from photographs. Scientific Reports, 11(1), 16395. https://www.nature.com/articles/s41598-021-95616-0

"To create this database, we utilized the TRY database (trait information) and the iNaturalist database (citizen science plant photographs). Based on the species names found in both databases, we linked the trait observations obtained from the TRY database (species-specific mean and standard deviation) with the plant photographs (iNaturalist). Based on the geocoordinates that comes with each plant photographs, we linked the ancillary predictors, which are derived from globally available raster data (WORLDCLIM, SOIL, VOD, MODIS). To state briefly, WORLDCLIM includes temperature and precipitation data, SOIL is the global soil grids dataset (interpolated products on various soil properties, such as sand content or pH value), MODIS is satellite data that measures optical reflectance of sun light (like a camera but with many wavelengths), while VOD represents data from a radar constellation that is sensitive to water content and biomass of plants. All these geodatasets are meant to serve as supporting information in addition to the plant photographs."