license: mit
pipeline_tag: image-feature-extraction
GreenHyperSpectra: A multi-source hyperspectral dataset for global vegetation trait prediction
This repository contains models and data associated with the paper GreenHyperSpectra: A multi-source hyperspectral dataset for global vegetation trait prediction.
GreenHyperSpectra introduces a pretraining dataset of real-world cross-sensor and cross-ecosystem hyperspectral samples. This dataset is designed to benchmark trait prediction using semi- and self-supervised methods. The work demonstrates how leveraging GreenHyperSpectra can lead to label-efficient multi-output regression models that outperform state-of-the-art supervised baselines, significantly improving the learning of spectral representations for plant trait prediction.
All code and data for this project are available at the official GitHub repository: https://github.com/GreenHyperSpectra/GreenHyperSpectra