Open Vocabulary Retrieval is an information retrieval paradigm in which a system matches queries to documents or data without being constrained to a predefined set of terms or categories. By leveraging dense neural representations and vision-language models, open vocabulary retrieval systems can handle arbitrary natural language queries and generalize to unseen concepts at inference time. Research challenges include retrieval accuracy on rare or compositional queries, efficient indexing at scale, and cross-modal retrieval across text, images, and other data types.
Satellite imagery is vast, high-resolution, and rich with information, but finding specific objects within it using natural language has remained a stubborn challenge. Open-SAT, developed by researchers at NEC Laboratories America and North South University, tackles this problem without retraining any models.