"Maize Seedling Detection Dataset (MSDD): A Curated High-Resolution RGB Dataset for Seedling Maize Detection and Benchmarking"

"Accurate maize seedling detection is crucial for modern precision agriculture, yet curated datasets for this purpose remain scarce. To address this gap, we introduce MSDD, a high-quality dataset of aerial images designed for maize seedling stand counting, which has broad applications in early-season crop monitoring, yield prediction, and in-field management decision-making. Stand counting helps determine how many plants have successfully germinated, enabling farmers to make timely adjustments such as replanting or modifying input applications. Traditional stand counting methods are labor- intensive and error-prone, whereas automated detection using computer vision offers significant improvements in efficiency and accuracy. MSDD comprises three classes \u2014 single plant, double plants, and triple plants \u2014 to ensure precise stand counting. This dataset includes aerial images captured under diverse conditions, featuring different growth stages, planting setups, soil colors, illuminations, camera angles, and planting densities, making it robust for real-world applications. Benchmarking results show that detection performs best during the V4\u2013V6 growth stages and under nadir camera views. Among the evaluated models, YOLO11 is the fastest, while YOLOv9 provides the highest accuracy for detecting single maize plants. While single plant detection achieves high precision up to 0.984 and recall up to 0.873, detecting double and triple plants remains challenging due to their rarity and anomalous appearance, often caused by planting errors. The imbalanced distribution of these categories contributes to reduced model accuracy in identifying multiple plants per stand. Despite these challenges, the fastest model, YOLO11, maintains efficient inference with an average process- ing time of 35 milliseconds per image, plus an additional 120 milliseconds for saving prediction bounding boxes and annotated images. MSDD provides a critical foundation for developing robust detection models that enhance stand counting precision, optimize resource allocation, and and supports real-time decision- making throughout the field season. This dataset marks a significant step toward the automation of agricultural monitoring, contributing to the advancement of precision agriculture."