Introduces YieldSAT, a large-scale multimodal benchmark for high-resolution crop yield prediction combining satellite imagery, weather, soil, and harvester data — enabling …
Studies how multimodal crop-yield models can be made more interpretable, examining modality contributions and model behavior rather than treating multimodal fusion as a black box.
Focuses on accurate sub-field yield prediction by combining multiple remote sensing and auxiliary modalities, highlighting the value of adaptive multimodal fusion for improving …
Presents spatial-context-aware multimodal fusion approaches for crop yield prediction, leveraging local neighborhood information to improve field and sub-field level predictions.
Evaluates how different data cleaning strategies for harvester yield maps affect the quality of sub-field crop yield predictions, revealing the sensitivity of ML models to …
Presents a broad empirical comparison of models and modality combinations for crop yield prediction, establishing a strong baseline for later work on multimodal fusion and …