Comparative Analysis of DenseNet121 and MobileNetV2 for Orange Disease and Health State Classification in Precision Agriculture

This study investigates the performance of the two state-of-the-art CNN architectures, namely DenseNet121 and MobileNetV2, considering the classification of conditions for oranges into classes such as Fresh, Canker, Blackspot, and Greening. Results showed that DenseNet121 was able to yield 97% accuracy, while its class-wise F1-score was 0.94, 0.93, 0.97, and 0.95. It gave excellent macro and weighted averages for precision, recall, and F1-score by averages of 0.96 and 0.95, respectively. In contrast, MobileNet V2 had an accuracy of 95% and performed with class-specific F1-scores of 0.93, 0.91, 0.95, and 0.99. For MobileNetV2, both macro and weighted averages were consistent at 0.95. While both have been putting up a strong performance, the overall accuracy and consistency are a wee bit more balanced in the case of DenseNet121. This analysis underlines the efficiency of both models in precision agriculture, with DenseNet121 tends to be slightly better in classifying orange diseases and health states.