Automated Seed Classification Using State-of-the-Art Techniques

The demand for efficient and accurate seed assessment is paramount in modern agriculture to ensure good crop yield. This work presents a system for automated seed assaying which utilizes advanced image recognition and deep learning algorithms to streamline the seed segregation process, addressing challenges associated with manual methods. Image analysis was carried out using Deep learning algorithms, enabling the classification as well as identification seeds based on various quality features such as shape, size, color, and surface features. Apart from accelerating the assessment process, this technique also enhances accuracy by minimizing human error and subjectivity. The proposed work employs Resnet50, MobileNetv2, DenseNet121, YOLOv5 and YOLOv8 models for classification of seeds. Among all these models, the YOLOv8 model gave better performance with an accuracy of 91%. The proposed work in this study has the potential to revolutionize seed quality control practices, ultimately contributing to increased agricultural productivity and sustainability.