Abstract
Abstract
Soybean is a key crop due to its nutritional, industrial, and agricultural benefits. Its production performance faces challenges such as improper storage, seed damage, and moisture imbalance. This paper aims to classify soybean seeds into five categories – healthy, damaged, immature, broken, and stained – using image processing and deep learning. Four classifiers DenseNet, MobileNet, ResNet, and EfficientNet were used to classify seeds on a dataset of 5513 soybean seed images. The best performance among the models used was obtained from DenseNet with an accuracy of 96% and an F1-score of 94%. The results also showed that all models achieved an AUC value above 0.99, indicating high performance of the models. This research highlights the potential of machine learning and artificial intelligence in optimising soybean classifiers.