Real-Time Maize Seed Quality Sorting Using YOLO11n

Seed quality is a major factor in farming, having direct implications on farm yield and profitability to farmers. Traditional maize seed sorting utilizes labor-intensive inspection by hand, time-consuming, variable, and fallible. With a view of overcoming these defects, we introduce an improved seed sorting system using the YOLO11n model, which is a good and lightweight object detection deep-learning model optimized for realtime application. We developed our own dataset consisting of 2,682 images of maize seed, uni formerly categorized under Good Seeds $(1,341)$ and Bad Seeds $(1,341)$. The YOLO11n model was then trained on the dataset to correctly classify seeds. Our model also had good recall, precision, and a mean Average Precision, indicating superior detection capability. Inference rate was optimized to 3.6 milliseconds per image, which is appropriate for real-time sorting within agricultural environments. Farmers are able to increase seed selection accuracy, lower labor expenses, and raise overall crop yields by using this automated seed sorting system. The suggested method ensures that only quality seeds are planted, resulting in improved germination rates and higher agricultural efficiency. Our method also offers scalability for application in seed processing plants, enabling progress of precision agriculture. Future development will involve expanding the model to accommodate several types of seeds and implementing it in automated seed sorting machines. The study acknowledges the potential of deep learning to transform conventional farming methods, increasing the efficiency and profitability of agriculture