Predicting Citrus Limon Maturity with Precision Using Transfer Learning

The traditional approach of assessing Citrus limon's ripeness level relies on subjective human judgment, resulting in sluggishness and potential errors. Given the large lemon production, an automated, speedy, and dependable system for predicting lemon conditions is essential. Our primary aim is to employ advanced methodologies to categorize lemons based on maturity level, surpassing the limitations of visual inspection. A vital determinant of ripeness is the lemon's color, which exhibits specific attributes in mature specimens. Leveraging transfer learning, our system utilizes pre-trained models to extract color features and forecast ripe lemons. Initially, we acquire lemon-related image samples and preprocess them to eliminate background interference. These images are then input into a Neural Network architecture, leveraging pre-training to refine the model's feature extraction efficiency. Furthermore, we meticulously test different maturity classes using pre-labeled samples, achieving a remarkable prediction accuracy rate exceeding 96%. Our proposed system demonstrates superior significance compared to alternative classification methods, as substantiated by extensive experimentation and evaluation.