PhytoScope: A Public Benchmark Dataset for Multi-Crop Disease Detection and Localization

PhytoScope: A Large-Scale Benchmark Dataset for Multi-Crop Disease Detection and Localization Description PhytoScope is a publicly available large-scale benchmark dataset developed to advance research in automated crop disease detection, localization, and intelligent agricultural monitoring using deep learning and computer vision techniques. The dataset has been designed to address the growing demand for comprehensive, real-world agricultural image datasets that support object detection rather than conventional image classification. Unlike many publicly available plant disease datasets that primarily consist of laboratory-captured leaf images with single-label classification annotations, PhytoScope contains disease instances collected under diverse field environments, enabling researchers to develop and evaluate robust object detection models suitable for real-world agricultural applications. The dataset contains more than 20,000 manually annotated field images collected from 25 economically important crop species, covering 105 disease and healthy object categories. Every disease instance has been carefully annotated using the YOLO object detection format, making the dataset directly compatible with modern object detection frameworks such as YOLOv8, YOLOv9, YOLOv10, YOLOv11, YOLOv12, RT-DETR, Faster R-CNN, EfficientDet, DETR, and other deep learning architectures. The primary objective of PhytoScope is to provide an open benchmark for developing accurate, efficient, and reproducible crop disease detection systems capable of operating under realistic agricultural conditions. Dataset Highlights Large-scale benchmark dataset for crop disease detection More than 20,000 annotated field images 25 major crop species 105 disease and healthy object categories Real-world field images collected under natural environmental conditions YOLO object detection annotations Public benchmark for reproducible research Suitable for both academic and industrial research Compatible with modern deep learning frameworks Crop Categories The dataset includes images from the following crops: Apple Bean Bitter Gourd Bottle Gourd Cauliflower Chilli Corn Cotton Cucumber Dragon Fruit Eggplant Ginger Grape Jackfruit Jute Lychee Mango Orange Papaya Potato Rice Strawberry Tea Tomato Watermelon These crops represent a diverse collection of fruits, vegetables, cereals, fiber crops, horticultural plants, and plantation crops that are widely cultivated across different agricultural regions.