GEOBench-VLM
收藏资源简介:
GEOBench-VLM数据集由穆罕默德·本·扎耶德人工智能大学创建,旨在评估视觉语言模型在地理空间任务中的表现。该数据集包含超过10,000个问题,涵盖了多种任务,如场景理解、对象计数、视觉定位、细粒度分类和时间分析。数据集的创建过程结合了自动化和手动验证,确保了数据的高质量。该数据集主要应用于环境监测、城市规划和灾害管理等领域,旨在解决地理空间数据分析中的复杂问题。
The GEOBench-VLM dataset was developed by Mohamed bin Zayed University of Artificial Intelligence, with the core objective of evaluating the performance of vision-language models (VLMs) on geospatial tasks. This dataset contains over 10,000 questions covering a diverse set of tasks including scene understanding, object counting, visual grounding, fine-grained classification, and temporal analysis. The dataset's construction combines automated processes and manual validation to ensure high data quality. It is primarily applied in fields such as environmental monitoring, urban planning, and disaster management, aiming to address complex problems in geospatial data analysis.

- 1GEOBench-VLM: Benchmarking Vision-Language Models for Geospatial Tasks穆罕默德·本·扎耶德人工智能大学 · 2024年



