REOBench
收藏资源简介:
REOBench是一个全面的基准测试,旨在评估地球观测基础模型的鲁棒性。我们的基准测试系统地评估了广泛的流行基础模型的鲁棒性,包括基于掩码图像建模、对比学习和大型语言模型的最先进模型。REOBench专注于高分辨率光学遥感图像,这些图像广泛应用于城市规划、灾害响应等现实世界应用中。我们在六种广泛研究的遥感图像理解任务上进行了实验,涵盖了视觉中心和视觉语言任务,在十二种类型的扰动下进行。这些包括外观基础的损坏(例如噪声、模糊、雾霾)和几何扭曲(例如旋转、缩放、平移),应用不同的严重程度以模拟现实环境和传感器引起的挑战。
REOBench is a comprehensive benchmark designed to evaluate the robustness of Earth Observation fundamental models. Our benchmark systematically assesses the robustness of a wide range of popular fundamental models, including state-of-the-art models based on masked image modeling, contrastive learning, and large language models. REOBench focuses on high-resolution optical remote sensing images, which are extensively used in real-world applications such as urban planning and disaster response. We conducted experiments on six widely studied remote sensing image understanding tasks, covering both visual and visual language tasks, under twelve types of disturbances. These include appearance-based damage (e.g., noise, blur, haze) and geometric distortion (e.g., rotation, scaling, translation), applied at different severity levels to simulate challenges posed by real-world environments and sensors.
REOBench数据集概述
数据集简介
- 名称:REOBench
- 全称:Benchmarking Robustness of Earth Observation Foundation Models
- 目的:评估地球观测基础模型的鲁棒性
- 数据类型:高分辨率光学遥感图像
- 应用领域:城市规划、灾害响应等
关键特性
- 评估范围:覆盖6种广泛研究的遥感图像理解任务
- 任务类型:视觉中心任务和视觉语言任务
- 扰动类型:12种扰动类型(包括外观和几何失真)
- 扰动级别:多种严重程度级别
数据集结构
- 语义分割:使用mmsegmentation进行实验
- 目标检测:使用mmrorate进行实验
- 分类:提供分类实验细节
- 视觉语言任务:包括图像描述、视觉问答和视觉定位
获取与使用
- 下载地址:https://huggingface.co/datasets/xiang709/REOBench
- 加载代码: python from datasets import load_dataset fw = load_dataset("xiang709/REOBench", streaming=True)
许可信息
- 许可证类型:CC-BY-4.0
- 许可链接:https://creativecommons.org/licenses/by/4.0/deed.en
相关项目
- VRSBench:多功能视觉语言基准数据集
- CLAIR:基于GPT的自动图像描述评估工具
引用信息
bibtex @inproceedings{livrsbench, title={VRSBench: A Versatile Vision-Language Benchmark Dataset for Remote Sensing Image Understanding}, author={Li, Xiang and Ding, Jian and Elhoseiny, Mohamed}, booktitle={The Thirty-eight Conference on Neural Information Processing Systems Datasets and Benchmarks Track} }
致谢
- 基础数据集:AID、DIOR、VRSBench
- 使用工具:mmdetection、mmsegmentation




