A Multimodal UAV Dataset for Rice Yield Prediction in the Vietnam Mekong Delta

This repository contains a multimodal UAV dataset for rice yield prediction in the Vietnam Mekong Delta (VMD), including 184 plot-level RGB UAV images acquired at the flowering stage, grain-yield measurements, vegetation-index features, agronomic variables, EffNet-derived image features, and source code associated with the study. Vegetation-index features include GNDVI, NDRE, LCI, SIPI2, CIre, and NDYI, extracted across six phenological stages: jointing (JT), booting (BT), heading (HD), flowering (FL), milk (ML), and dough (DG). Deep spatial features were generated using the EffNet (EfficientNet-B0) architecture through a five-fold out-of-fold (OOF) framework. The repository includes complete and selected vegetation-index datasets, multimodal datasets, FL-stage datasets, EffNet feature representations, plot-level UAV images, and source code for image preprocessing, feature extraction, machine-learning models, ensemble model, and SHAP analysis. The repository is provided to support reproducible research on UAV-based rice yield prediction, multimodal machine learning, explainable artificial intelligence, and precision agriculture.