tirtho149/SAGE

SAGE是一个用于作物病害诊断的可扩展代理基础评估数据集，包含约1,010,311张图像，格式为Parquet分片。数据来自公开植物病理学数据集和专家策划的集合，涵盖多种作物物种（如大豆、玉米、芒果）和病害条件标签。每行数据包括原始JPEG/PNG图像字节、作物种类、病害标签和原始文件名。该数据集适用于图像分类和零样本图像分类任务，主要用于农业领域的代理人工智能和视觉语言模型研究。注意数据集存在地理偏见（主要覆盖北美温带作物）和类别不平衡（大豆、玉米、小麦占主导），不推荐用于医疗诊断或未经验证的地理部署。SAGE: Scalable Agentic Grounded Evaluation for Crop Disease Diagnosis is a dataset containing ~1,010,311 images in Parquet shard format, aggregated from public plant-pathology datasets and expert-curated collections. It includes columns for raw image bytes, crop species (e.g., Soybean, Corn, Mango), disease/condition labels, and original filenames. The dataset is designed for image-classification and zero-shot-image-classification tasks, focusing on crop disease and agriculture applications, with tags like agentic AI and vision-language models. It highlights responsible AI considerations such as geographic bias (primarily North American temperate-zone crops) and class imbalance (dominated by Soybean, Corn, Wheat), and is not recommended for medical diagnosis or deployment in unseen geographies without adaptation.