Work

Field Rice Disease and Pest Dataset

Year: 2026
Type: dataset
Abstract: This dataset contains 9,296 high-resolution images of rice diseases and pests acquired under unconstrained field conditions. It was specifically curated to train and evaluate computer vision models (such as YOLO and State-Space Models) in complex agricultural environments involving varied illumination, severe occlusion, and visually ambiguous lesions.Dataset Statistics & Split:Total Images: 9,296Training Set: 7,436 images (acquired from 96 independent capture sessions)Validation Set: 1,860 images (acquired from 24 independent capture sessions)Note: The dataset was rigorously partitioned at the "capture-session" level to prevent temporal and spatial data leakage between the training and validation subsets.Categories (6 Classes): 0: Rice Leaf Roller 1: Rice Water Weevil 2: Rice Leafhopper 3: Bacterial Blight 4: Brown Spot 5: Rice BlastAnnotation Format: The annotations are provided in standard YOLO flat-text format (.txt). Each text file contains bounding box coordinates normalized as:Mandatory Citation Policy: If you use this dataset in your academic research, industrial applications, or publications, you MUST cite our foundational manuscript. Please cite the following preprint (or its subsequent peer-reviewed journal version):Zhu, H., Luo, X., Zeng, Y., Tan, W., Zeng, Q., & Qian, H. (2025). Lightweight State-Space YOLO for Field Rice Disease and Pest Detection: Statistical Evaluation, Robustness Analysis, and Edge-NPU Deployment. Research Square. DOI: https://doi.org/10.21203/rs.3.rs-9928737/v1
Source: Figshare
Author: Hairong Zhu
Language: English
Subfield: Plant Science
Open Access status: green
OpenAlex ID: W7213030854
Cites:
Cited by:
Related to: 10