This is a page that curates AI-related papers published worldwide. All content here is summarized using Google Gemini and operated on a non-profit basis. Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.
This paper presents AgroMind, a comprehensive benchmark for evaluating the performance of large-scale multimodal models (LMMs) in agricultural remote sensing (RS). To overcome the limitations of existing benchmarks, which often lack dataset diversity and oversimplified task design, AgroMind encompasses four task dimensions and 13 task types: spatial perception, object understanding, scene understanding, and scene inference. By integrating eight public datasets and one private farmland dataset, we constructed a high-quality evaluation set consisting of 27,247 QA pairs and 19,615 images. Evaluating 20 open-source LMMs and four closed-source models on AgroMind, we found significant performance differences, particularly in spatial inference and fine-grained recognition, with some top-performing LMMs outperforming human performance. AgroMind provides a standardized evaluation framework for agricultural RS, exposing the domain-specific limitations of LMMs and highlighting important challenges for future research. Data and code are available at https://rssysu.github.io/AgroMind/ .
