Multi-class Classification of Coconut Leaf Disease using Fine-tuned ResNet50 Model

Coconuts are an important staple crop and vital commodity in Southeast Asia. There are various diseases causing coconut tree health degradation including leaflets, CCI caterpillars and Flaccidity which seriously threaten crop health and productivity. To address these challenges, deep-learning models have proven to be powerful means of making predictions about various diseases. For, the classification of coconut leaf diseases, the proposed model uses a ResNet50 architecture, implemented with both Adam and SGD optimizers, to reach impressive accuracies of 98.7% and 97.9%, respectively, at epoch 40. This points to the fact that modern deep-learning techniques are highly robust, and represent a great leap forward in coconut leaf disease classification research. The study goes into depth about the subtle connection between generations and model quality, showing that an epoch value of 40 is best for combining efficiency with effectiveness. A value of epoch 10 has produced sub-optimal results in both Adam and SGD optimizers. Reaching its best results at epoch 40, the proposed strategy achieves F1-Scores of 0.9562 and 0.942 for Adam and SGD optimizers, respectively. This demonstrates the model's ability to differentiate between various diseased coconut leaves. Additionally, the flexible ResNet50 model presented in the proposed work is effectively capable of multi-class classification of coconut leaf diseases. It can also serve in classifying the other crop diseases with improved training sets. This research provides key insights into the application of deep learning toward a more accurate and effective prediction about crop diseases--which is especially important for Southeast Asian coconut agriculture.