FLORO: A Multimodal Geospatial Foundation Model for Ecological Remote Sensing Across Sensors and Scales
Researchers have introduced FLORO, a multimodal geospatial foundation model designed for ecological remote sensing applications. Unlike many existing models that require massive datasets and fixed sensor configurations, FLORO is trained on a diverse yet smaller corpus and incorporates availability-aware inputs to handle varying sensor data. The model demonstrated strong transferability across different image types and resolutions on the PANGAEA benchmark, achieving competitive results in segmentation, scene classification, and regression tasks. AI
IMPACT FLORO offers a new approach to foundation models for remote sensing, potentially improving ecological analysis with diverse and limited data.