Automated Lemon Quality Detection using DenseNet121: A Deep Learning Approach

Detection of lemon quality is an important process in agricultural and commercial sectors, optimizing the product standards and minimizing waste. This study explores the application of the DenseNet121 deep learning model for the classification of lemons into two categories: Bad Quality and Good Quality. Preprocessing of a dataset of 2,076 images from Kaggle was done using advanced techniques like resizing, normalization, CLAHE, and data augmentation to facilitate extraction of features and development of generalization of models. On the test set, the model was fine-tuned, and the accuracy was found to be 96%. The performance on all evaluation metrics, such as precision, recall, and F1-score, has always stayed at a very high level in both classes. Further analysis of the model using the confusion matrix revealed its reliability in reducing misclassifications. By automating the detection of lemon quality using deep learning, this study demonstrates the potential in automating processes in agriculture and supply chain management while also showcasing an efficient and scalable solution in the real world.