We integrate Hyperspectral (point spectra) and RGB images from herbarium specimens to build single-modal and multimodal classification models and evaluate the performance of taxonomic species identifications at the level of family, order, and genus.
This work was done as part of FloraPalooza 2026 see https://github.com/Imageomics/FloraPalooza-2026
Each branch contains a different data modality (rgb, hyperspectral, scalograms from hyperspectral) with corresponding data utils, models, and evaluation code.